Wall crack detection system based on image recognition

Abstract

The invention discloses a wall crack detection system based on image recognition, which relates to the technical field of building monitoring and comprises an image enhancement denoising module, a geometric feature extraction module, a structure health prediction module, a crack classification rating module, a crack extension monitoring module, a repair suggestion generation module and a crack detection verification module. According to the wall crack detection system based on image recognition, through adoption of crack geometric feature extraction, a structure health prediction model based on a graph convolutional network and a crack expansion monitoring module combining time sequence analysis and an optical flow method, the precision and real-time performance of wall crack detection are remarkably improved; trend prediction and self-adaptive feedback of crack expansion are realized, an efficient and intelligent solution is provided for health management of a building structure, and potential structural failure is effectively prevented.

Description

Technical Field

[0001] The invention relates to the technical field of building monitoring, and in particular to a wall crack detection system based on image recognition. Background Art

[0002] In the process of building maintenance and management, wall cracks are an important monitoring object, because the expansion of cracks may affect the structural safety of buildings. At present, crack detection systems based on image recognition have been gradually applied to the field of building monitoring. Traditional crack detection systems mainly rely on manual inspection or simple image processing techniques such as edge detection and threshold segmentation. However, these traditional methods have great limitations: manual inspection is time-consuming and laborious, and is easily affected by subjective factors; simple image processing technology cannot accurately identify cracks in complex backgrounds and is easily disturbed by environmental factors such as light and noise. In addition, existing crack monitoring systems are usually limited to static detection of cracks, lack real-time monitoring of dynamic crack expansion and long-term trend prediction functions, and cannot effectively prevent building structure failure. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a wall crack detection system based on image recognition to solve the problems raised in the above background technology.

[0004] To achieve the above-mentioned object, the present invention provides the following technical solutions: a wall crack detection system based on image recognition, comprising an image enhancement and denoising module, a geometric feature extraction module, a structural health prediction module, a crack classification and rating module, a crack extension monitoring module, a repair suggestion generation module and a crack detection and verification module; The image enhancement and denoising module performs adaptive image enhancement and denoising on the image of the wall crack, and automatically adjusts parameters according to the overall brightness and noise level of the image; The geometric feature extraction module, based on the multi-scale geometric feature extraction method of convolutional neural network, extracts information on the length, width, curvature and depth of the crack from different levels, automatically identifies the starting point and end point of the crack, and calculates the overall geometric shape of the crack; The structural health prediction module uses a structural health prediction model of a graph convolutional network based on the geometric characteristics and historical data of the cracks to predict the future health status of the wall according to the expansion trend of the cracks; The crack classification and rating module classifies wall cracks according to their morphology and development stage, and rates the degree of damage to the wall; The crack expansion monitoring module combines time series analysis and optical flow method to capture the crack expansion speed and direction changes in real time under long-term monitoring, and integrates environmental sensor data of temperature and humidity based on multi-sensor fusion to provide multi-dimensional crack development analysis; The repair suggestion generation module generates a personalized repair plan based on crack characteristics and building type combined with a building material database; The crack detection verification module verifies whether the cracks detected in the image have deep structural impact based on ultrasonic detection and infrared imaging technology.

[0005] Further optimizing the technical solution, when the image enhancement and denoising module performs adaptive image enhancement and denoising, the interference in the image is filtered out by the adversarial generative network based on deep learning, and the visual features of the crack area are enhanced through contrast and brightness adjustment strategies; At the same time, the noise is processed based on a multi-scale filter to make the edge image of the crack clearer after processing.

[0006] Further optimizing the technical solution, in the geometric feature extraction module, for wall crack detection, the geometric features of the cracks, such as length, width, curvature, and depth, show different details and forms at different scales. The multi-scale geometric feature extraction method captures the global and local information of the cracks by extracting the features of the cracks at different resolutions or scales. Multi-scale geometric feature extraction methods include: The lower convolutional layers capture finer geometric details, such as the edges and textures of cracks; High-level convolutional layers focus on the global structure and morphology of cracks; By combining the features of different convolutional layers, the geometric information of cracks at various scales is extracted; Based on multi-scale technology, cracks of different sizes and shapes are dealt with to ensure the robustness and accuracy of crack feature extraction.

[0007] To further optimize the technical solution, the structural health prediction model in the structural health prediction module includes the following contents: Define the graphical model of building structure and cracks, assuming that the geometric characteristics of the cracks and the influencing factors of the environment and external forces are represented as a graph ( ),in( ) is a node set, representing the characteristic points or key geometric positions of the crack, ( ) is an edge set, which represents the topological relationship between different feature points, such as the connection between the starting point and the end point of a crack, or the association between a crack and a structural stress point; Perform graph convolution operations. The operation formula of the graph convolution network is as follows: ; in, ( ) is the first ( ) layer node feature matrix, which represents the crack feature information of the current layer; ( )equal( ), represents the adjacency matrix of the graph, ( ) represents the connection relationship in the graph, ( ) is the identity matrix, indicating self-connection; ( ) is the degree matrix of the graph; ( ) is the first ( ) layer weight matrix, used to learn features; ( ) is a nonlinear activation function such as ReLU, which is used to introduce nonlinear characteristics; Through graph convolution operations, each layer of the graph convolution network propagates and aggregates crack geometric features on the graph structure, capturing the relationship between the crack node and its neighboring nodes.

[0008] Further optimizing the technical solution, the structural health prediction model constructs an objective function in the output layer of the graph convolutional network to evaluate the structural health risk when predicting crack expansion and the structural health status of the wall. Assuming that we want to predict the future moment ( ),the health status of the wall ( ), the prediction task is expressed as: ; in, ( ) is the final feature matrix after multiple layers of convolution; ( ) is the geometric characteristics of the crack at the current moment, including length, width, and depth; ( ) are external mechanical parameters, such as external loads, temperature changes, and environmental factors such as humidity; By inputting the features into a deep learning-based regression or classification model, the structural health status ( )’s predicted value; Based on time series modeling, the expansion trend of cracks in the time dimension is captured. Assuming that at the current moment ( ) is related to the previous crack state and is optimized by the recursive formula: ; The recursive formula represents the current moment ( )'s health status prediction depends not only on the current features ( ), but also affected by historical status ( ) and accurately predict future wall health by considering historical evolution trends.

[0009] To further optimize the technical solution, the crack classification and rating module adopts a deep learning classification algorithm to identify different types of cracks, such as horizontal cracks, vertical cracks, and reticular cracks, through a pre-trained model; The cracks are rated for wall damage based on their width, depth and length.

[0010] To further optimize this technical solution, in the crack extension monitoring module, Based on time series analysis, the data of cracks at different times are modeled to detect the long-term trend and periodic changes of crack expansion, including the use of historical data to predict future crack behavior. It is suitable for situations where crack expansion is slow but the cumulative impact is large; Based on the optical flow method, the pixel changes in the crack area are calculated by analyzing the movement of each pixel in the continuous frame images to detect the displacement direction and speed of the crack, and track the changes of the crack in real time in a short period of time. It is suitable for situations where the crack expands rapidly or the dynamic external force is obvious.

[0011] To further optimize the present technical solution, the crack extension monitoring module also has a built-in adaptive feedback mechanism driven by AI, which automatically adjusts the monitoring frequency according to the real-time changes of the cracks and triggers an early warning mechanism when the cracks reach a certain critical point.

[0012] To further optimize the technical solution, the adaptive feedback mechanism includes adaptive monitoring frequency adjustment, parameter optimization and automatic adjustment, and early warning mechanism, which specifically includes the following contents: Adaptive monitoring frequency adjustment: Dynamically adjust the monitoring frequency according to the crack expansion speed; when the crack expansion is slow, if the historical data and real-time monitoring results show that the crack expansion is relatively stable, reduce the monitoring frequency to reduce unnecessary resource consumption; when the crack expansion accelerates, the crack expansion speed is detected to increase, and the monitoring frequency is automatically increased to ensure more detailed detection at critical moments; Parameter optimization and automatic adjustment: Dynamically optimize system parameters based on current monitoring data. If the width or depth of the crack changes significantly, automatically adjust the sensitivity of the detection, giving priority to features with large changes. Dynamically adjust the parameter weights of the detection algorithm through the temperature and humidity data obtained by the sensor to make the crack extension analysis conform to the actual situation. Early warning mechanism: When the expansion trend of the cracks reaches a certain critical value, an early warning is automatically triggered; the threshold of the early warning mechanism is not fixed. AI determines the early warning critical point suitable for the current scenario based on different types of building materials, crack expansion patterns and external environmental data.

[0013] To further optimize this technical solution, in the crack detection verification module, Infrared imaging technology identifies temperature anomalies inside the wall, thereby determining whether there are invisible deep cracks; Ultrasonic testing determines the depth of cracks and the degree of internal damage through echo analysis inside the material.

[0014] Compared with the prior art, the present invention provides a wall crack detection system based on image recognition, which has the following beneficial effects: This wall crack detection system based on image recognition has significantly improved the accuracy and real-time performance of wall crack detection by extracting crack geometric features, adopting a structural health prediction model based on a graph convolutional network, and a crack expansion monitoring module that combines time series analysis and optical flow method. It has also achieved trend prediction and adaptive feedback of crack expansion, providing an efficient and intelligent solution for building structure health management and effectively preventing potential structural failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a structural schematic diagram of a wall crack detection system based on image recognition proposed by the present invention; Figure 2 A schematic flow chart of a geometric feature extraction method in a wall crack detection system based on image recognition proposed by the present invention; Figure 3 A schematic diagram of a structural health prediction model in a wall crack detection system based on image recognition proposed by the present invention; Figure 4 This is a schematic diagram of an adaptive feedback mechanism in a wall crack detection system based on image recognition proposed by the present invention. DETAILED DESCRIPTION

[0016] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example: See also Figure 1 ,A wall crack detection system based on image recognition includes an image enhancement and denoising module, a geometric feature extraction module, a structural health prediction module, a crack classification and rating module, a crack extension monitoring module, a repair suggestion generation module, and a crack detection verification module.

[0018] In practical applications, the image of wall cracks may have poor detection results due to insufficient light, dust coverage, or differences in wall material. In order to improve the recognition accuracy of cracks, adaptive image enhancement and denoising are performed on the image of wall cracks, and the parameters are automatically adjusted according to the overall brightness and noise level of the image.

[0019] In this embodiment, when the image enhancement and denoising module performs adaptive image enhancement and denoising, the interference in the image is filtered out based on the deep learning-based adversarial generative network, and the visual features of the crack area are enhanced through contrast and brightness adjustment strategies; At the same time, the noise is processed based on a multi-scale filter to make the edge image of the crack clearer after processing, which is convenient for subsequent analysis.

[0020] The geometric feature extraction module, based on a multi-scale geometric feature extraction method of a convolutional neural network, extracts information on the length, width, curvature and depth of the crack from different levels, automatically identifies the starting point and end point of the crack, and calculates the overall geometric shape of the crack.

[0021] like Figure 2 As shown, in this embodiment, in the geometric feature extraction module, for wall crack detection, the geometric features of the cracks, such as length, width, curvature, and depth, show different details and forms at different scales. The multi-scale geometric feature extraction method captures the global and local information of the cracks by extracting the features of the cracks at different resolutions or scales; Multi-scale geometric feature extraction methods include: The lower convolutional layers capture finer geometric details, such as the edges and textures of cracks; High-level convolutional layers focus on the global structure and morphology of cracks; By combining the features of different convolutional layers, the geometric information of cracks at various scales is extracted; Based on multi-scale technology, cracks of different sizes and shapes are dealt with to ensure the robustness and accuracy of crack feature extraction.

[0022] The structural health prediction module uses a graph convolutional network structural health prediction model based on the geometric characteristics and historical data of the cracks to predict the future health status of the wall according to the expansion trend of the cracks. Compared with the traditional rule-based prediction method, it can more accurately predict the possible development path of the cracks through the structural characteristics of the graph. In addition, it can capture the slight changes in the cracks through comparative analysis of crack images in multiple time periods, and then predict the deterioration rate of the wall.

[0023] like Figure 3 As shown, in this embodiment, the structural health prediction model in the structural health prediction module includes the following contents: Define the graphical model of building structure and cracks, assuming that the geometric characteristics of the cracks and the influencing factors of the environment and external forces are represented as a graph ( ),in( ) is a node set, representing the characteristic points or key geometric positions of the crack, ( ) is an edge set, which represents the topological relationship between different feature points, such as the connection between the starting point and the end point of a crack, or the association between a crack and a structural stress point; Perform graph convolution operations. The operation formula of the graph convolution network is as follows: ; in, ( ) is the first ( ) layer node feature matrix, which represents the crack feature information of the current layer; ( )equal( ), represents the adjacency matrix of the graph, ( ) represents the connection relationship in the graph, ( ) is the identity matrix, indicating self-connection; ( ) is the degree matrix of the graph; ( ) is the first ( ) layer weight matrix, used to learn features; ( ) is a nonlinear activation function such as ReLU, which is used to introduce nonlinear characteristics; Through graph convolution operations, each layer of the graph convolution network propagates and aggregates crack geometric features on the graph structure, capturing the relationship between the crack node and its neighboring nodes.

[0024] The structural health prediction model constructs an objective function in the output layer of the graph convolutional network to evaluate the structural health risk when predicting crack expansion and the structural health status of the wall. Assuming that we want to predict the future moment ( ),the health status of the wall ( ), the prediction task is expressed as: ; in, ( ) is the final feature matrix after multiple layers of convolution; ( ) is the geometric characteristics of the crack at the current moment, including length, width, and depth; ( ) are external mechanical parameters, such as external loads, temperature changes, and environmental factors such as humidity; By inputting the features into a deep learning-based regression or classification model, the structural health status ( ). This value can be a score (such as a risk score between 0 and 1), or a categorical label (such as "healthy", "dangerous", "needs repair", etc.).

[0025] Based on time series modeling, the expansion trend of cracks in the time dimension is captured. Assuming that at the current moment ( ) is related to the previous crack state and is optimized by the recursive formula: ; The recursive formula represents the current moment ( )'s health status prediction depends not only on the current features ( ), but also affected by historical status ( ) and accurately predict future wall health by considering historical evolution trends.

[0026] The model usage process is as follows: Data input: Input the geometric characteristics of the cracks (such as length, width, depth) and the spatial distribution of the cracks to construct a map ( ). At the same time, input environmental data (such as temperature, humidity, external force, etc.) and historical crack development.

[0027] Graph construction: Based on the geometric characteristics of the cracks and the wall structure, a graph model is constructed to define the crack feature points and the relationships between them to form a representation of nodes and edges.

[0028] Model training: The model is trained using supervised learning methods through historical crack expansion data. The model will continuously adjust the weight matrix ( ) to minimize the prediction error.

[0029] Health status prediction: Input the current crack characteristics and external conditions into the trained model to obtain the current moment ( )’s structural health prediction results.

[0030] Optimization and iteration: Dynamically adjust the model based on the actual crack expansion situation. The introduction of the time series model can better predict the future crack expansion trend and continuously optimize the prediction accuracy through data feedback.

[0031] The model can handle complex crack topology and identify the spatial relationship between crack features, which can capture the complex dynamics of crack expansion more accurately than traditional linear models. It integrates geometric features, historical data, external environmental factors and structural mechanical properties to form a multi-dimensional input, making the prediction more comprehensive. The prediction ability can be automatically optimized through the continuous accumulation of historical data, especially with good adaptability when dealing with different building structures or environmental conditions.

[0032] The crack classification and rating module classifies wall cracks according to their morphology and development stage, and rates the degree of damage to the wall.

[0033] In this embodiment, the crack classification and rating module uses a deep learning classification algorithm to identify different types of cracks, such as horizontal cracks, vertical cracks, and reticular cracks, through a pre-trained model; The cracks are rated for wall damage based on their width, depth and length.

[0034] The crack expansion monitoring module combines time series analysis and optical flow method to capture the crack expansion speed and direction changes in real time under long-term monitoring, and integrates environmental sensor data of temperature and humidity based on multi-sensor fusion to provide multi-dimensional crack development analysis.

[0035] In this embodiment, in the crack extension monitoring module, Time series analysis is based on modeling data of cracks at different times to detect long-term trends and periodic changes in crack expansion, including using historical data to predict future crack behavior. It is suitable for situations where crack expansion is slow but the cumulative impact is large. Time series analysis can help the system identify chronic problems of crack expansion, such as crack expansion caused by long-term loads or external environmental pressures, and provide a reference for predicting future health conditions.

[0036] Based on the optical flow method, the pixel changes in the crack area are calculated by analyzing the movement of each pixel in the continuous frame images to detect the displacement direction and speed of the crack, and track the changes of the crack in real time in a short period of time. It is suitable for situations where the crack expands rapidly or the dynamic external force is obvious.

[0037] In summary, the combination of time series analysis and optical flow method forms a complete solution that combines short-term real-time response and long-term trend analysis. Optical flow method is responsible for rapid detection and feedback of crack expansion in the short term, while time series analysis provides in-depth analysis of long-term crack expansion trends. The two combine environmental data (such as temperature, humidity and other sensor information) to provide a comprehensive and three-dimensional monitoring perspective for crack expansion.

[0038] For example, when the optical flow method detects that the crack is expanding faster in the short term, the system can verify through the time series model whether this change is consistent with the long-term trend or whether there is a sudden impact from the external environment (such as a sudden temperature change). Combining this dual analysis method, the system can provide a more reliable early warning of crack expansion.

[0039] The crack extension monitoring module also has a built-in AI-driven adaptive feedback mechanism, which automatically adjusts the monitoring frequency according to the real-time changes of the cracks and triggers an early warning mechanism when the cracks reach a certain critical point.

[0040] like Figure 4 As shown, the adaptive feedback mechanism includes adaptive monitoring frequency adjustment, parameter optimization and automatic adjustment, and early warning mechanism, which specifically includes the following contents: Adaptive monitoring frequency adjustment: Dynamically adjust the monitoring frequency according to the crack expansion speed; when the crack expansion is slow, if the historical data and real-time monitoring results show that the crack expansion is relatively stable, reduce the monitoring frequency to reduce unnecessary resource consumption; when the crack expansion accelerates, the crack expansion speed is detected to increase, and the monitoring frequency is automatically increased to ensure more detailed detection at critical moments; Parameter optimization and automatic adjustment: Dynamically optimize system parameters based on current monitoring data. If the width or depth of the crack changes significantly, automatically adjust the sensitivity of the detection, giving priority to features with large changes. Dynamically adjust the parameter weights of the detection algorithm through the temperature and humidity data obtained by the sensor to make the crack extension analysis conform to the actual situation. Early warning mechanism: When the crack expansion trend reaches a certain critical value, an early warning is automatically triggered; the threshold of the early warning mechanism is not fixed, and AI determines the early warning critical point suitable for the current scenario based on different types of building materials, crack expansion patterns, and external environmental data. The early warning mechanism can provide multiple levels of alarms based on the severity of crack expansion. For example, a level 1 alarm can indicate that crack expansion is accelerating and further monitoring is recommended; a level 2 alarm may indicate that cracks may threaten the safety of the building structure and immediate action is recommended.

[0041] The repair suggestion generation module generates a personalized repair plan based on crack characteristics and building types in combination with a building material database.

[0042] In this example, the best repair method is selected based on historical repair data and environmental conditions, such as grouting repair, crack sealant, surface covering or structural reinforcement. The module can also predict the effects of different repair plans and provide a comparative analysis of cost and time to help users choose the best repair strategy.

[0043] The crack detection verification module verifies whether the cracks detected in the image have deep structural impact based on ultrasonic detection and infrared imaging technology.

[0044] In this embodiment, in the crack detection verification module, Infrared imaging technology identifies temperature anomalies inside the wall, thereby determining whether there are invisible deep cracks; Ultrasonic testing determines the depth of cracks and the degree of internal damage through echo analysis inside the material.

[0045] The beneficial effects of the present invention are: This wall crack detection system based on image recognition has significantly improved the accuracy and real-time performance of wall crack detection by extracting crack geometric features, adopting a structural health prediction model based on a graph convolutional network, and a crack expansion monitoring module that combines time series analysis and optical flow method. It has also achieved trend prediction and adaptive feedback of crack expansion, providing an efficient and intelligent solution for building structure health management and effectively preventing potential structural failures.

[0046] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0047] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various 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 wall crack detection system based on image recognition, characterized in that: It includes image enhancement and denoising module, geometric feature extraction module, structural health prediction module, crack classification and rating module, crack extension monitoring module, repair suggestion generation module and crack detection and verification module; The image enhancement and denoising module performs adaptive image enhancement and denoising on the image of the wall crack, and automatically adjusts parameters according to the overall brightness and noise level of the image; The geometric feature extraction module, based on the multi-scale geometric feature extraction method of convolutional neural network, extracts information on the length, width, curvature and depth of the crack from different levels, automatically identifies the starting point and end point of the crack, and calculates the overall geometric shape of the crack; The structural health prediction module uses a structural health prediction model of a graph convolutional network based on the geometric characteristics and historical data of the cracks to predict the future health status of the wall according to the expansion trend of the cracks; The crack classification and rating module classifies wall cracks according to their morphology and development stage, and rates the degree of damage to the wall; The crack expansion monitoring module combines time series analysis and optical flow method to capture the crack expansion speed and direction changes in real time under long-term monitoring, and integrates temperature and humidity environmental sensor data based on multi-sensor fusion to provide multi-dimensional crack development analysis; The repair suggestion generation module generates a personalized repair plan based on crack characteristics and building type combined with a building material database; The crack detection verification module verifies whether the cracks detected in the image have deep structural impact based on ultrasonic detection and infrared imaging technology.

2. A wall crack detection system based on image recognition according to claim 1, characterized in that: When the image enhancement and denoising module performs adaptive image enhancement and denoising, interference in the image is filtered out based on a deep learning-based adversarial generative network, and visual features of the crack area are enhanced through contrast and brightness adjustment strategies; At the same time, the noise is processed based on a multi-scale filter to make the edge image of the crack clearer after processing.

3. A wall crack detection system based on image recognition according to claim 1, characterized in that: In the geometric feature extraction module, for wall crack detection, the geometric features of the cracks, such as length, width, curvature, and depth, show different details and forms at different scales. The multi-scale geometric feature extraction method captures the global and local information of the cracks by extracting the features of the cracks at different resolutions or scales. Multi-scale geometric feature extraction methods include: The lower convolutional layers capture finer geometric details, such as the edges and textures of cracks; High-level convolutional layers focus on the global structure and morphology of cracks; By combining the features of different convolutional layers, the geometric information of cracks at various scales is extracted; Based on multi-scale technology, cracks of different sizes and shapes are dealt with to ensure the robustness and accuracy of crack feature extraction.

4. A wall crack detection system based on image recognition according to claim 1, characterized in that: The structural health prediction model in the structural health prediction module includes the following contents: Define the graphical model of building structure and cracks, assuming that the geometric characteristics of the cracks and the influencing factors of the environment and external forces are represented as a graph ( ),in( ) is a node set, representing the characteristic points or key geometric positions of the crack, ( ) is an edge set, which represents the topological relationship between different feature points, such as the connection between the starting point and the end point of a crack, or the association between a crack and a structural stress point; Perform graph convolution operations. The operation formula of the graph convolution network is as follows: ； in, ( ) is the first ( ) layer node feature matrix, which represents the crack feature information of the current layer; ( )equal( ), represents the adjacency matrix of the graph, ( ) represents the connection relationship in the graph, ( ) is the identity matrix, indicating self-connection; ( ) is the degree matrix of the graph; ( ) is the first ( ) layer weight matrix, used to learn features; ( ) is a nonlinear activation function such as ReLU, which is used to introduce nonlinear characteristics; Through graph convolution operations, each layer of the graph convolution network propagates and aggregates crack geometric features on the graph structure, capturing the relationship between the crack node and its neighboring nodes.

5. A wall crack detection system based on image recognition according to claim 4, characterized in that: The structural health prediction model constructs an objective function in the output layer of the graph convolutional network to evaluate the structural health risk when predicting crack expansion and the structural health status of the wall. Assuming that we want to predict the future moment ( ),the health status of the wall ( ), the prediction task is expressed as: ； in, ( ) is the final feature matrix after multiple layers of convolution; ( ) is the geometric characteristics of the crack at the current moment, including length, width, and depth; ( ) are external mechanical parameters, such as external loads, temperature changes, and environmental factors such as humidity; By inputting the features into a deep learning-based regression or classification model, the structural health status ( )’s predicted value; Based on time series modeling, the expansion trend of cracks in the time dimension is captured. Assuming that at the current moment ( ) is related to the previous crack state and is optimized by the recursive formula: ； The recursive formula represents the current moment ( )'s health status prediction depends not only on the current features ( ), but also affected by historical status ( ) and accurately predict future wall health based on historical evolution trends.

6. A wall crack detection system based on image recognition according to claim 1, characterized in that: The crack classification and rating module adopts a deep learning classification algorithm to identify different types of cracks, such as horizontal cracks, vertical cracks, and reticular cracks, through a pre-trained model; The cracks are rated for wall damage based on their width, depth and length.

7. A wall crack detection system based on image recognition according to claim 1, characterized in that: In the crack extension monitoring module, Based on time series analysis, the data of cracks at different times are modeled to detect the long-term trend and periodic changes of crack expansion, including the use of historical data to predict future crack behavior. It is suitable for situations where crack expansion is slow but the cumulative impact is large; Based on the optical flow method, the pixel changes in the crack area are calculated by analyzing the movement of each pixel in the continuous frame images to detect the displacement direction and speed of the crack, and track the changes of the crack in real time in a short period of time. It is suitable for situations where the crack expands rapidly or the dynamic external force is obvious.

8. The wall crack detection system based on image recognition according to claim 1, characterized in that: The crack extension monitoring module also has a built-in AI-driven adaptive feedback mechanism, which automatically adjusts the monitoring frequency according to the real-time changes of the cracks and triggers an early warning mechanism when the cracks reach a certain critical point.

9. A wall crack detection system based on image recognition according to claim 8, characterized in that: The adaptive feedback mechanism includes adaptive monitoring frequency adjustment, parameter optimization and automatic adjustment, and early warning mechanism, which specifically includes the following contents: Adaptive monitoring frequency adjustment: Dynamically adjust the monitoring frequency according to the crack expansion speed; when the crack expansion is slow, if the historical data and real-time monitoring results show that the crack expansion is relatively stable, reduce the monitoring frequency to reduce unnecessary resource consumption; when the crack expansion accelerates, the crack expansion speed is detected to increase, and the monitoring frequency is automatically increased to ensure more detailed detection at critical moments; Parameter optimization and automatic adjustment: Dynamically optimize system parameters based on current monitoring data. If the width or depth of the crack changes significantly, automatically adjust the sensitivity of the detection, giving priority to features with large changes. Dynamically adjust the parameter weights of the detection algorithm through the temperature and humidity data obtained by the sensor to make the crack extension analysis conform to the actual situation. Early warning mechanism: When the crack expansion trend reaches a certain critical value, the early warning is automatically triggered; The threshold of the early warning mechanism is not fixed. AI determines the early warning critical point suitable for the current scenario based on different types of building materials, crack expansion patterns and external environmental data.

10. The wall crack detection system based on image recognition according to claim 1, characterized in that: In the crack detection verification module, Infrared imaging technology identifies temperature anomalies inside the wall, thereby determining whether there are invisible deep cracks; Ultrasonic testing determines the depth of cracks and the degree of internal damage through echo analysis inside the material.

Images