Method for measuring instantaneous concrete crack width change

Abstract

The invention discloses a method for measuring instantaneous concrete crack width change, which comprises the following steps: selecting a target concrete crack, and pasting a calibration object beside the target concrete crack; an industrial camera is erected in advance at the position away from the target concrete by the preset length, and a crack video corresponding to a target concrete crack is collected through the industrial camera; training the crack segmentation model by adopting an improved deep learning network to obtain a trained crack segmentation model; and performing crack width change analysis on the crack video based on the trained crack segmentation model and a pre-integrated full-process algorithm to obtain crack width change data corresponding to the target concrete crack and a visual file thereof. According to the method, the high-frame-rate industrial camera and the improved deep learning network are combined, the instantaneous dynamic change of the concrete crack width is accurately measured, the accuracy of crack segmentation is effectively improved, the change trend of the crack width is visually displayed, and the efficiency and accuracy of bridge structure health monitoring are improved.

Description

Technical Field

[0001] The present application relates to the technical field of structural health monitoring, and in particular to a method for measuring instantaneous changes in concrete crack width. Background Art

[0002] In the field of bridge structural health monitoring, ensuring the safety and stability of bridges during their service is of vital importance. Cracks are one of the main factors affecting the service life of bridges. Their appearance not only affects the appearance of bridges, but also may provide a channel for harmful substances inside the concrete, causing steel corrosion, thus affecting the safety and service life of bridges. Therefore, crack detection has become an important task in bridge maintenance.

[0003] Traditional bridge crack detection methods mainly rely on manual visual inspection, which has the problems of low efficiency and low accuracy. It also consumes a lot of manpower and economic resources, resulting in instability in the detection results. With the development of technology, bridge crack detection methods based on artificial intelligence have gradually become a hot topic of research. These methods include using machine vision to enhance the analysis of crack images to calculate parameters, or using deep learning algorithms to automatically identify cracks and calculate parameters. However, most of these methods only focus on the static observation of cracks, that is, obtaining crack width information by taking a crack image, but lack effective means of observing the dynamic changes of cracks within a certain period of time. Therefore, there is an urgent need for a method that can capture and analyze the instantaneous dynamic changes of crack width.

[0004] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the invention

[0005] The embodiment of the present application provides a method for measuring instantaneous changes in concrete crack width, which can more quickly and accurately obtain the width change trend of concrete cracks at instantaneous high frequency, and provide new data support for structural health assessment.

[0006] In order to solve the above technical problems, the present application provides a method for measuring instantaneous concrete crack width change, comprising:

[0007] Select a target concrete crack and paste a calibration object next to the target concrete crack;

[0008] An industrial camera is pre-installed at a preset length from the target concrete, and a crack video corresponding to the crack of the target concrete is collected by the industrial camera;

[0009] An improved deep learning network is used to train the crack segmentation model to obtain a trained crack segmentation model;

[0010] Based on the trained crack segmentation model and the pre-integrated full-process algorithm, crack width change analysis is performed on the crack video to obtain crack width change data corresponding to the target concrete crack and its visualization file.

[0011] Further, in some embodiments of the present application, the step of pre-setting an industrial camera at a preset length from the target concrete and collecting a crack video corresponding to the crack of the target concrete by the industrial camera includes:

[0012] Setting up a camera bracket at a preset length from the target concrete, placing a working camera on the camera bracket, and connecting the working camera to a computer;

[0013] The target concrete crack with the calibration object attached thereto is photographed by the working camera to obtain a calibrated concrete crack image;

[0014] The calibration object is removed, and the crack video corresponding to the target concrete crack is collected by the working camera.

[0015] Furthermore, in some embodiments of the present application, the improved deep learning network is used to train the crack segmentation model to obtain a trained crack segmentation model, including:

[0016] Select several crack images and their corresponding labels to construct a segmented image dataset, and preprocess the segmented image dataset;

[0017] Using an improved deep learning network as the crack segmentation model network;

[0018] The crack segmentation model network is trained based on the preprocessed segmented image data set to obtain a trained crack segmentation model.

[0019] Further, in some embodiments of the present application, the use of an improved deep learning network as a crack segmentation model network includes:

[0020] Use the residual network structure ResNet-50 as the backbone feature extraction network;

[0021] Adding an SE-Net module during the skip connection process of the backbone feature extraction network so that the backbone feature extraction network determines the importance of different channels by learning to assign different weights;

[0022] An SK-Attention module is added to the upsampling of the network to enable the network to perform multi-scale feature fusion.

[0023] Furthermore, in some embodiments of the present application, the integration method of the full-process algorithm includes:

[0024] Construct crack segmentation prediction algorithm, crack width calculation algorithm and result visualization algorithm;

[0025] According to the crack analysis process, the crack segmentation prediction algorithm, the crack width calculation algorithm and the result visualization algorithm are integrated into a corresponding full-process algorithm.

[0026] Further, in some embodiments of the present application, the crack width change analysis is performed on the crack video based on the trained crack segmentation model and the pre-integrated full-process algorithm to obtain the crack width change data corresponding to the target concrete crack and its visualization file, including:

[0027] The crack video is segmented by using the crack segmentation prediction algorithm in the full-process algorithm to obtain a plurality of frames of crack segmentation images;

[0028] The crack widths of the plurality of crack segmentation images are calculated respectively by the crack width calculation algorithm in the full-process algorithm to obtain corresponding crack width calculation results;

[0029] The crack width calculation result is visualized by the result visualization algorithm in the full-process algorithm to obtain crack width change data corresponding to the target concrete crack and its visualization file.

[0030] Furthermore, in some embodiments of the present application, the crack video is segmented by the crack segmentation prediction algorithm in the full-process algorithm to obtain a plurality of frames of crack segmentation images, including:

[0031] Performing frame-by-frame segmentation processing on the crack video to obtain a plurality of frames of crack segmentation images to be predicted;

[0032] The crack segmentation prediction processing is performed on the plurality of frames of crack segmentation images to be predicted using a pre-trained crack segmentation model to obtain a plurality of corresponding crack segmentation images.

[0033] Further, in some embodiments of the present application, the crack width calculation algorithm in the full-process algorithm is used to calculate the crack widths of the plurality of crack segmentation images to obtain corresponding crack width calculation results, including:

[0034] Performing skeleton extraction on the plurality of crack segmentation image frames to obtain a central skeleton line corresponding to the crack in each crack segmentation image frame;

[0035] An eight-direction calculation algorithm is used to search with the central skeleton line as the center to determine whether there are target pixel points in the crack outline in each direction and the number thereof;

[0036] The target pixel points in the diagonal direction are geometrically merged, and the minimum pixel length is selected as the crack pixel width to obtain the crack width calculation result corresponding to each frame of the crack segmentation image.

[0037] Further, in some embodiments of the present application, the crack width calculation result is visualized by the result visualization algorithm in the full-process algorithm to obtain the crack width change data corresponding to the target concrete crack and its visualization file, including:

[0038] Performing data analysis on the crack width calculation results by using a result visualization algorithm to obtain a data file of the maximum width and average width corresponding to each frame of the crack segmentation image;

[0039] Based on the data file of the maximum width and the average width corresponding to each frame of the crack segmentation image, a curve graph and a visual video showing the variation of the maximum width and the average width of the crack with the video time are generated.

[0040] Furthermore, in some embodiments of the present application, the method for generating the visualization video includes:

[0041] Converting the color mode of the crack segmentation image from the RGB mode to the HSV mode;

[0042] Determine the maximum crack width based on the data file of the maximum width and the average width corresponding to each frame of the crack segmentation image;

[0043] Calculating the color width of each skeleton point based on the maximum crack width;

[0044] Calculate the hue corresponding to each skeleton point based on the color width;

[0045] According to the hue, preset saturation and brightness corresponding to each skeleton point, the color mode of the crack segmentation image is converted from the HSV mode to the RGB mode to generate a corresponding visualization video.

[0046] The present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program as described in the steps of the above method.

[0047] The present application also provides a readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the method described above when the computer program is executed by a processor.

[0048] The present application provides a method for measuring instantaneous changes in concrete crack width, including: selecting a target concrete crack and pasting a calibration object next to the target concrete crack; pre-setting up an industrial camera at a preset length from the target concrete, and collecting a crack video corresponding to the target concrete crack through the industrial camera; using an improved deep learning network to train a crack segmentation model to obtain a trained crack segmentation model; based on the trained crack segmentation model and a pre-integrated full-process algorithm, performing crack width change analysis on the crack video to obtain crack width change data corresponding to the target concrete crack and its visualization file. The scheme for measuring instantaneous concrete crack width changes provided by the present application, by combining a high frame rate industrial camera and an improved deep learning network, can achieve accurate measurement of instantaneous dynamic changes in concrete crack width, effectively improve the accuracy of crack segmentation, and intuitively display the changing trend of crack width, thereby improving the efficiency and accuracy of bridge structure health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0050] Figure 1 Schematic diagram of the application environment of the method for measuring instantaneous concrete crack width change provided by the embodiment of the present application;

[0051] Figure 2 is a flow chart of a method for measuring instantaneous concrete crack width change provided in an embodiment of the present application;

[0052] Figure 3 is a schematic diagram of the structure of the calibration object provided in the embodiment of the present application;

[0053] Figure 4 It is a schematic diagram of the structure of the U-Net network provided in the embodiment of the present application;

[0054] Figure 5 It is a schematic diagram of the structure of the residual structure and upsampling provided in the embodiment of the present application;

[0055] Figure 6 It is a structural diagram of the SE-Net module provided in an embodiment of the present application;

[0056] Figure 7 is a schematic diagram of the structure of the SK-Attention module provided in an embodiment of the present application;

[0057] Figure 8 This is a schematic diagram of the improved network model for detecting and segmenting cracks provided in an embodiment of the present application;

[0058] Fig. 9 is a schematic diagram of crack width visualization provided by an embodiment of the present application;

[0059] Fig.10 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of systems and methods consistent with the examples detailed in the attached claims, or with some aspects of the present application.

[0061] It should be noted that, in this article, the terms "include", "comprises" or any other variants thereof are intended to cover descriptions such as non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined by their explanation in the specific embodiment or further combined with the context of the specific embodiment.

[0062] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] In the subsequent description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present application, and have no specific meanings. Therefore, "module", "component" or "unit" can be used in a mixed manner.

[0064] In order to ensure that the bridge structure can operate safely and stably during its service life, regular health monitoring and maintenance work must be carried out. There are many factors that affect the service life of bridges, among which cracks are one of the main factors. Its appearance will not only directly affect the aesthetics of the overall structure of the bridge, but also provide a channel for harmful substances inside the concrete, causing the steel bars inside to be easily corroded, thus affecting the safety and service life of the bridge. Therefore, bridge crack detection is one of the important tasks at present. Manual visual inspection is the main means of bridge crack detection. However, this method is not only inefficient and inaccurate, but also requires a lot of money, resulting in instability in the detection results. And the current method is only a single static observation of bridge cracks, that is, taking a crack image, aiming to obtain crack width information more accurately; for a certain period of time, the dynamic change of crack width cannot be observed.

[0065] In order to solve the above technical problems, an embodiment of the present application provides a method for measuring instantaneous changes in concrete crack width, which can more quickly and accurately obtain the width change trend of concrete cracks at instantaneous high frequency, providing new data support for structural health assessment.

[0066] Figure 1 FIG. 1 is an application environment diagram of a method for measuring instantaneous concrete crack width change in an embodiment. Figure 1 , the method for measuring instantaneous concrete crack width change can be applied to a system for measuring instantaneous concrete crack width change. Among them, the system for measuring instantaneous concrete crack width change can include a terminal 110 and a server 120. The terminal 110 and the server 120 are connected through a network. The terminal 110 can be a desktop terminal or a mobile terminal. The mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The server 120 can be implemented as an independent server or a server cluster composed of multiple servers. The terminal 110 is used to select a target concrete crack and paste a calibration object next to the target concrete crack; an industrial camera is pre-installed at a preset length from the target concrete, and a crack video corresponding to the target concrete crack is collected through the industrial camera; an improved deep learning network is used to train a crack segmentation model to obtain a trained crack segmentation model; crack width change analysis is performed on the crack video based on the trained crack segmentation model and the pre-integrated full-process algorithm to obtain crack width change data corresponding to the target concrete crack and its visualization file.

[0067] See also Figure 2 , Figure 2 : is a flow chart of a method for measuring instantaneous concrete crack width change provided by an embodiment of the present application. The method for measuring instantaneous concrete crack width change provided by an embodiment of the present application may specifically include the following steps:

[0068] S1. Select the target concrete crack and paste the calibration object next to the target concrete crack;

[0069] Specifically, for step S1, select the target concrete crack to be measured this time, that is, determine the specific crack location to be monitored. Paste the calibration object next to the crack. The actual size of the calibration object is known and is used for subsequent calculation of pixel scale to ensure the measurement accuracy of the crack width. Select the target concrete crack and paste the printed square calibration sticker with a side length of 4 mm, such as Figure 3 As shown, it should be flat and without wrinkles and pasted on the edge of the crack. It should not cover the crack to be measured. The color of the square calibration object should be black.

[0070] S2. Pre-install an industrial camera at a preset length from the target concrete, and collect crack videos corresponding to the target concrete cracks through the industrial camera;

[0071] Specifically, for step S2, a high frame rate industrial camera is set up at a preset length from the target concrete crack, and the distance between the camera and the crack is kept constant to ensure measurement consistency. The crack video is recorded by the high frame rate industrial camera to record the dynamic changes of the crack.

[0072] S3. Using an improved deep learning network to train the crack segmentation model to obtain a trained crack segmentation model;

[0073] Specifically, for step S3, a crack original image and segmented image dataset are constructed, and the data is preprocessed to match the network input requirements. An improved U-Net network is used as the training segmentation model network, which uses ResNet-50 as the backbone feature extraction network, and adds SE-Net and SK-Attention modules to improve performance.

[0074] S4. Analyze the crack width change of the crack video based on the trained crack segmentation model and the pre-integrated full-process algorithm to obtain the crack width change data corresponding to the target concrete crack and its visualization file;

[0075] Specifically, for step S4, the crack segmentation prediction algorithm, the crack width calculation algorithm and the result visualization algorithm are integrated to form a full-process algorithm. The crack video is analyzed using the trained crack segmentation model and the full-process algorithm to obtain crack width change data and its visualization file.

[0076] It can be seen that the method for measuring instantaneous changes in concrete crack width provided in this embodiment reduces the workload of manual monitoring and improves monitoring efficiency through automated crack width change analysis; utilizes high frame rate cameras and deep learning technology to more accurately measure crack width and provide more reliable structural health assessment data; the improved deep learning network and full-process algorithm can more accurately identify and segment cracks, improving the accuracy and robustness of crack detection; uses visualization technology to make changes in crack width more intuitive, making it easier for engineers and decision makers to quickly understand and respond; the provided crack width dynamic change data provides new data support for health assessments of structures such as bridges, helps prevent structural failures, and ensures public safety.

[0077] Further, in some embodiments, step S2 of "pre-setting an industrial camera at a preset length from the target concrete, and collecting a crack video corresponding to the crack of the target concrete through the industrial camera" may specifically include:

[0078] S21. Set up a camera bracket at a preset length from the target concrete, place the working camera on the camera bracket, and connect the working camera to the computer;

[0079] S22. Taking a picture of the target concrete crack with the calibration object pasted thereon by the working camera to obtain a concrete crack image with calibration;

[0080] S23. Remove the calibration object and collect crack videos corresponding to the target concrete cracks through the working camera.

[0081] Specifically, for step S2, a camera bracket is set up at a preset length from the target concrete crack to ensure that the camera maintains a constant distance and angle during the shooting process to obtain consistent image data. The industrial camera is placed on the camera bracket and connected to a computer for real-time transmission and processing of image data. First, take a picture of the target concrete crack with the calibration object attached to it, and obtain the concrete crack image with the calibration object for subsequent pixel scale calculation. After removing the calibration object, keep the camera position and posture unchanged, collect the crack video of the target concrete crack, and record the dynamic changes of the crack.

[0082] This embodiment ensures that the collected image data has a high degree of consistency by maintaining a constant distance and angle between the camera and the crack, which is beneficial to the subsequent crack width measurement and analysis; it realizes real-time acquisition and transmission of crack videos, making crack monitoring more real-time and able to promptly detect dynamic changes in cracks; it obtains pixel scales through crack images with calibration objects, and combines with high frame rate industrial cameras to improve the accuracy of crack width measurement.

[0083] Furthermore, in some embodiments, step S3 of "using an improved deep learning network to train a crack segmentation model to obtain a trained crack segmentation model" may specifically include:

[0084] S31. Select several crack images and their corresponding labels, construct a segmented image dataset, and preprocess the segmented image dataset;

[0085] S32. Using an improved deep learning network as a crack segmentation model network;

[0086] S33. Train the crack segmentation model network based on the preprocessed segmented image data set to obtain a trained crack segmentation model.

[0087] Specifically, for step S3, the main thing is to build an improved deep learning network (for example, a U-Net network structure). First, select several crack images and their corresponding labels to build a segmented image dataset. The dataset is resized and image enhanced to ensure that the input data format matches the network input requirements. Then, the improved deep learning network is used as the crack segmentation model network. The network uses the residual network structure ResNet-50 as the backbone feature extraction network, adds the SE-Net module in the jump connection process, and adds the SK-Attention module in the upsampling for multi-scale feature fusion. Finally, the crack segmentation model network is trained based on the preprocessed segmented image dataset to obtain a trained crack segmentation model.

[0088] This embodiment improves the accuracy of crack detection by using an improved deep learning network, especially a residual network structure and an attention mechanism. The application of data enhancement and multimodal data fusion technology enhances the model's generalization ability for crack images under different conditions. Transfer learning reduces the time and computing resources required for model training, making model training more efficient. Model optimization and compression technology make the model more suitable for deployment on resource-constrained edge devices, improving the practical application value of the model. Through model optimization, data processing efficiency is improved, making the analysis of crack width changes faster and more accurate.

[0089] Further, in some embodiments, step S32 “using the improved deep learning network as the crack segmentation model network” may specifically include:

[0090] Use the residual network structure ResNet-50 as the backbone feature extraction network;

[0091] Add the SE-Net module in the skip connection process of the backbone feature extraction network so that the backbone feature extraction network can determine the importance of different channels by learning to assign different weights;

[0092] The SK-Attention module is added to the upsampling of the network to enable the network to perform multi-scale feature fusion.

[0093] Specifically, for step S32, an improved U-Net network structure is constructed, such as Figure 4 As shown in Figure 1, it includes an encoder and a decoder. The ResNet-50 model is used as the main component of the encoder. Figure 5 As shown in , after passing through 5 different residual layers, each layer outputs the extracted feature map, such as Figure 6 As shown in , the SE-Net channel attention mechanism is used to enhance the features. Then it is fused with the upsampled feature map corresponding to the same number of channels, and then Figure 7 The SK-Attention module shown captures information of different scales to improve the network's focus on extracting crack features, and ultimately achieves the binary classification problem of cracks and background.

[0094] It is understandable that this embodiment uses the residual network structure ResNet-50 as the backbone feature extraction network to alleviate the gradient vanishing problem and improve the network's deep learning ability. The SE-Net module is added in the jump connection process of the backbone feature extraction network to determine more important channels by learning to assign different weights and improve network performance. The SK-Attention module is added in upsampling so that the network can capture information of different scales and increase the accuracy of the segmentation model.

[0095] By integrating SE-Net and SK-Attention modules, this embodiment enables the network to more accurately identify crack areas and improve the accuracy of crack segmentation. The ResNet-50 structure helps the network learn deeper features, while the SE-Net and SK-Attention modules enhance the importance of features and the ability to capture multi-scale information. The residual network structure effectively alleviates the gradient vanishing problem, allowing the network to be trained deeper and improving the performance of the model. Through multi-scale feature fusion technology and cross-domain adaptability, the model can be better generalized to different data sets and environmental conditions.

[0096] Further, in some embodiments, step S33 "training the crack segmentation model network based on the preprocessed segmented image data set to obtain a trained crack segmentation model" specifically includes the following process:

[0097] The processed images and labels are divided into training sets and validation sets in a certain ratio and input into the network for training. For example, 7,000 crack images and corresponding labels are selected from the open source dataset on the Internet, and the dataset is resized and image enhanced to ensure that the input data format matches the network input requirements. The ratio of training set, validation set and test set is set to 8:1:1, the initial batch is set to 4, and the rounds are set to 200. In order to optimize the network parameters, the Adam optimizer is used to prevent overfitting. The initial learning rate is set to 0.001. The loss function is Dice Loss, and it is put into the network model for training. Adjust the network parameters according to the loss function.

[0098] Use Dice Loss as the loss function, set the initial learning rate, batch size, and smoothing parameters. As training progresses, gradually reduce the learning rate, and adjust the smoothing parameters and batch size based on the performance of the validation set. Evaluate model performance regularly and use the early stopping mechanism to prevent overfitting. The loss function formula is as follows:

[0099]

[0100] Where p represents the probability value predicted by the model; t represents the binary value (0 or 1) of the true label; p i , t i Represent the predicted value and true value of the i-th pixel respectively.

[0101] (5) After training, in order to test the accuracy of the model, the following indicators are used to evaluate the model: precision (P), recall (R) and F1 score (F1). Precision refers to the proportion of samples predicted as positive that are actually positive. Recall measures the proportion of actual positive samples found among samples predicted as positive by the model. F1 score is an indicator for comprehensively evaluating the performance of the classification model, which is the harmonic mean of precision and recall. The calculation formulas for each evaluation indicator are as follows:

[0102]

[0103] Where P is the precision, R is the recall, F1 is the F1 value, TP is the true positive example, FP is the false positive example, and FN is the false negative example.

[0104] According to the test set test, the evaluation index of the improved network of the present invention is P 91.2%, R 85.3%, F1 88.1%, which is a significant improvement compared with the original network P 82.4%, R 77.6%, F1 79.9%. Figure 8 As shown, the white part represents the crack and the black part represents the background.

[0105] Furthermore, in some embodiments, the integration of the full-process algorithm includes:

[0106] Construct crack segmentation prediction algorithm, crack width calculation algorithm and result visualization algorithm;

[0107] According to the crack analysis process, the crack segmentation prediction algorithm, crack width calculation algorithm and result visualization algorithm are integrated into the corresponding full-process algorithm.

[0108] Specifically, the integration method of the full-process algorithm provided in this embodiment is as follows: construct a crack segmentation prediction algorithm, use the trained crack segmentation model to segment the crack video frame by frame, and obtain the crack segmentation image; develop a crack width calculation algorithm, perform pixel scale calculation and crack width calculation on the crack segmentation image, and obtain the width data of the crack in each frame. Create a result visualization algorithm to convert the crack width data into a visualization format, such as a color mapping video, to intuitively display the changes in the crack width. The above three algorithms are integrated into a full-process algorithm to realize the automated processing from video input to the output of crack width dynamic change data.

[0109] Among them, algorithms can be optimized to reduce computing time, such as through parallel processing or the use of more efficient data structures. GPUs can be used to accelerate computing, especially in the deep learning and image processing stages. Cloud computing can be used for large-scale data processing and storage, while edge computing can be used to achieve rapid on-site analysis and response.

[0110] The integration of the full-process algorithm in this embodiment realizes the automated processing from video input to the output of data on dynamic changes in crack width, reducing manual intervention and improving efficiency. The result visualization algorithm enables the changes in crack width to be intuitively displayed through color mapping and other methods, which is easy to understand and analyze. Algorithm optimization and GPU acceleration significantly improve the data processing speed, making real-time monitoring possible. The combination of cloud computing and edge computing provides flexibility and scalability in data processing to meet the needs of projects of different scales.

[0111] Further, in some embodiments, step S4 "performing crack width change analysis on the crack video based on the trained crack segmentation model and the pre-integrated full-process algorithm to obtain crack width change data corresponding to the target concrete crack and its visualization file" may specifically include:

[0112] S41. Segment the crack video using the crack segmentation prediction algorithm in the full process algorithm to obtain several frames of crack segmentation images;

[0113] S42. Calculating the crack widths of several frames of crack segmentation images respectively by using the crack width calculation algorithm in the full process algorithm to obtain corresponding crack width calculation results;

[0114] S43. The crack width calculation results are visualized by the result visualization algorithm in the full-process algorithm to obtain crack width change data corresponding to the target concrete crack and its visualization file.

[0115] Specifically, for step S4, the filmed crack video is imported into the full-process algorithm, and the video is first segmented through the model trained by the segmentation network. The video is mainly frame extracted and each frame is segmented. Secondly, the pixel scale is calculated. In this example, the pixel scale is 0.0387. Then, each segmented frame image is sent to the width calculation algorithm to calculate the crack width according to the eight-direction method. Finally, the crack width corresponding to each skeleton point of each crack is compared, and the maximum and average values ​​are calculated to export the data file that changes over time. Set the color change speed and brightness to 1, the saturation to 0.8, calculate and draw the color mapping corresponding to the crack width, and export the visualization video. The video screenshot is shown as follows. Fig. 9 In this embodiment, the maximum value of the maximum crack width change is 0.36 mm, and the change range is about 0.02 mm.

[0116] Further, in some embodiments, step S41 of "segmenting the crack video using the crack segmentation prediction algorithm in the full-process algorithm to obtain a plurality of crack segmentation images" may specifically include:

[0117] The crack video is segmented frame by frame to obtain several frames of crack segmentation images to be predicted;

[0118] The crack segmentation prediction processing is performed on several frames of crack segmentation images to be predicted using a pre-trained crack segmentation model to obtain several corresponding frames of crack segmentation images.

[0119] Specifically, for step S41, a crack segmentation prediction algorithm is established, that is, based on the trained segmentation model, a crack video is imported, and segmentation prediction is performed frame by frame. The crack video file is read into the computer system using video processing software or programming libraries (such as OpenCV, FFmpeg). Make sure that the video file format is compatible with the processing software or library, such as common video formats such as MP4 and AVI. Extract images frame by frame from the imported video file. Each frame of the image represents a moment of the video, usually in the form of a separate image file or an image object in memory. Each frame can be read sequentially using a loop structure, for example, using the cv2.VideoCapture class and .read() method of OpenCV. Perform necessary preprocessing operations on each frame of the image to improve the accuracy of the segmentation model. Preprocessing may include resizing the image, normalizing, enhancing contrast, etc. The preprocessing steps should be consistent with the preprocessing steps used when training the deep learning model. Load the trained crack segmentation model. This usually involves reading the model weights and architecture from the file system, and deep learning frameworks such as TensorFlow, PyTorch, etc. can be used. Make sure the model has been properly trained and performs well on the crack segmentation task. Input each frame of the image into the loaded crack segmentation model and make a prediction. The model will output the crack segmentation result for each frame of the image, usually a binary image where the crack area is marked as white (or 1) and the non-crack area is marked as black (or 0). Post-process the segmentation results output by the model, such as morphological operations (opening, closing operations) to remove noise, or apply thresholding to further optimize the segmentation effect. The post-processing steps help improve the accuracy and robustness of crack detection. Store the segmentation results of each frame, which can be saved as an image file or stored in a data structure in memory for subsequent crack width calculation and analysis. Timestamps can be attached to the segmentation results of each frame to analyze the changes in crack width over time.

[0120] It can be understood that this embodiment provides an algorithm capable of processing real-time video streams so as to instantly detect and analyze crack changes as the cracks develop; introduces more advanced image segmentation techniques, such as a semantic segmentation network based on deep learning, to improve the accuracy of crack segmentation; combines other types of sensor data (such as infrared, sound waves) to assist crack segmentation and improve the segmentation effect in complex environments; and dynamically adjusts the segmentation threshold according to the content of the video frame to adapt to different lighting conditions and crack characteristics.

[0121] The real-time video stream processing capability provided by this embodiment makes crack detection more timely, which helps to quickly respond to structural health issues; by enhancing the crack segmentation algorithm, the accuracy of crack segmentation is improved, and false detections and missed detections are reduced; multimodal data fusion technology improves the crack segmentation effect under complex lighting and background conditions; adaptive segmentation threshold adjustment enables the algorithm to adapt to different environmental conditions, thereby improving the robustness of the algorithm.

[0122] Further, in some embodiments, step S42 of "calculating the crack widths of several frames of crack segmentation images respectively by the crack width calculation algorithm in the full process algorithm to obtain corresponding crack width calculation results" may specifically include:

[0123] Perform skeleton extraction on several frames of crack segmentation images to obtain the central skeleton line corresponding to the crack in each frame of crack segmentation image;

[0124] An eight-direction calculation algorithm is used to search with the central skeleton line as the center to determine whether there are target pixel points in the crack outline in each direction and the number of target pixel points;

[0125] The target pixel points in the diagonal direction are geometrically merged, and the minimum pixel length is selected as the crack pixel width to obtain the crack width calculation result corresponding to each frame of the crack segmentation image.

[0126] Specifically, for step S42, a width calculation algorithm is established, and pixel scale calculation is first performed using the following calculation formula:

[0127]

[0128] Where a represents the side length of the actual calibration object square, N represents the number of pixels in the calibration object square, and d is the pixel scale, that is, one pixel represents the actual distance.

[0129] Secondly, the crack width is calculated using the eight-direction calculation method, the steps are as follows:

[0130] (1) By extracting the skeleton of the binary crack, the central skeleton line of the crack is obtained.

[0131] (2) With these skeleton pixels as the center, search in eight directions from 0° to 360° at intervals of 45°, to determine whether there are pixels in the crack contour in each direction and to count the number of crack pixels in that direction.

[0132] (3) The pixel points in the diagonal direction are geometrically merged, and the minimum pixel length is selected as the crack pixel width. The actual crack width calculation formula is:

[0133] w=d×p(6)

[0134] Where d is the pixel scale, p is the crack pixel width, and w is the actual crack width.

[0135] In addition, more advanced image analysis techniques can be introduced, such as image analysis methods based on machine learning, to improve the accuracy of crack width calculations; automated crack feature recognition technology can be developed to automatically identify the starting point, end point, and branch of cracks to improve the accuracy of crack width calculations. Time series analysis or machine learning models can be integrated to predict the trend of crack width changes over time and provide forward-looking data for structural health monitoring.

[0136] This embodiment improves the accuracy of crack width measurement and ensures the reliability of data through skeleton extraction and eight-direction calculation algorithm; the automated crack feature recognition technology can more accurately identify the key features of cracks and improve the accuracy of crack analysis; the multi-scale crack analysis technology enables the algorithm to adapt to cracks of different widths and depths, improving the applicability of the algorithm; the parameters of the crack width calculation algorithm are adaptively adjusted to make the calculation results more adapted to the actual crack morphology and environmental conditions, improving the accuracy of the calculation results; by calculating the crack width in real time and predicting the change trend, the real-time and dynamic nature of structural health monitoring is enhanced, which helps to timely discover and respond to structural problems.

[0137] Further, in some embodiments, step S43 "visualizing the crack width calculation result through the result visualization algorithm in the full process algorithm to obtain the crack width change data corresponding to the target concrete crack and its visualization file" may specifically include:

[0138] The crack width calculation results are analyzed by the result visualization algorithm to obtain the data files of the maximum width and average width corresponding to each frame of the crack segmentation image;

[0139] Based on the data files of the maximum width and average width corresponding to each frame of the crack segmentation image, a curve graph and a visualization video showing the changes of the maximum width and average width of the crack with the video time are generated.

[0140] Specifically, for step S43, a data output algorithm is established, including outputting a data file of the maximum width and average width corresponding to each frame of the crack; outputting a curve chart showing the maximum width and average width of the crack changing with the video time; and outputting a visual video, which converts the crack width information into the color depth on the video through a specific color mapping method. The darker the color, the larger the crack width in the area, and vice versa.

[0141] In addition, interactive visualization tools can be developed to allow users to adjust color mapping parameters as needed to better understand changes in crack width; the visualization results of crack width can be combined with augmented reality technology to provide field workers with more intuitive crack analysis; using 3D modeling technology, crack width data can be converted into a 3D view to provide a more comprehensive crack analysis; combined with other sensor data, such as stress, humidity, etc., data fusion visualization can be performed to provide more comprehensive structural health information.

[0142] This embodiment uses a result visualization algorithm to convert complex crack width data into intuitive graphics and videos, thereby improving the comprehensibility of the data; the integration of interactive visualization tools and augmented reality technology enables field staff to understand and analyze crack conditions more intuitively, thereby improving the convenience of field operations; three-dimensional visualization and data fusion visualization provide more comprehensive crack analysis, which helps to deeply understand the nature and impact of cracks; the real-time visualization update interface enables monitoring personnel to instantly see changes in crack width, thereby improving the real-time nature of monitoring; and by providing a variety of visualization options, the user experience is improved, so that non-professional users can also understand and use crack monitoring data.

[0143] Furthermore, in some embodiments, the method of generating a visualization video includes:

[0144] Convert the color mode of the crack segmentation image from RGB mode to HSV mode;

[0145] Determine the maximum crack width based on the data file of the maximum width and the average width corresponding to each frame of the crack segmentation image;

[0146] Calculate the color width of each skeleton point based on the maximum crack width;

[0147] Calculate the hue corresponding to each skeleton point based on the color width;

[0148] According to the hue, preset saturation and brightness corresponding to each skeleton point, the color mode of the crack segmentation image is converted from HSV mode to RGB mode to generate the corresponding visualization video.

[0149] Specifically, for the color mapping method of the visual video, the calculation steps are as follows:

[0150] (1) In order to facilitate the calculation of the hue corresponding to each crack width and reduce the complexity of the calculation, the RGB mode of the image is converted to the HSV color mode. The starting hue hs of the color gradient is set to green, and the ending hue he is set to red. In order to show the distinction between different crack widths, the cs parameter is also set to control the speed of color change. The larger the value, the faster the color changes. The color changes between different crack widths are more obvious.

[0151] (2) According to the maximum crack width w obtained in the measured video max , calculate the color intensity c of each skeleton point i :

[0152] c i =w / w max (7)

[0153] (3) Calculate the hue corresponding to each skeleton point as:

[0154] hue=hs+c i ×(he-hs)×cs(8)

[0155] (4) Finally, the saturation and brightness are set according to the actual situation, converted to RGB mode, the color map of the crack width is output, and the visualization video is exported.

[0156] In addition, an adaptive color mapping algorithm can be developed to dynamically adjust the color mapping according to the range of changes in the crack width to improve the visual effect under different crack conditions; combine other relevant crack data (such as crack depth, temperature changes, etc.) to perform multi-dimensional data fusion visualization to provide a more comprehensive crack analysis; allow users to customize visualization parameters such as color mapping range, saturation, brightness, etc. to meet the needs and preferences of different users; create an interactive visualization interface where users can zoom in, zoom out, rotate the view, and select different data layers for display through interactive operations.

[0157] This embodiment enhances the visual recognition of crack width through the conversion of HSV color mode and the calculation of hue, making the change of crack width more intuitive; the adaptive color mapping algorithm and user-defined visualization parameters improve the flexibility of data display to adapt to different analysis needs; the interactive visualization interface improves the user experience, allowing users to explore and analyze data according to their own needs; multi-dimensional data fusion visualization enhances the depth of crack analysis and provides more comprehensive structural health information; high-quality visualization videos are easy to share and report, which helps to convey complex crack analysis results to decision makers with non-technical backgrounds.

[0158] In summary, the method for measuring instantaneous concrete crack width changes provided in this embodiment, firstly, uses a high frame rate camera to obtain high-frequency crack width change videos; secondly, the residual network structure ResNet-50, SE-Net and SK-Attention attention mechanism are used to improve the deep learning network, which can more accurately identify and segment crack images. Furthermore, a color mapping algorithm is established to realize the mapping form of crack width changes to color depth, and more intuitively perceive the instantaneous changes in crack width. A set of full-process algorithms is integrated to realize the automated processing from video input to crack width dynamic change data output. Finally, only a simple video input is required to obtain the crack width dynamic change data over time, which greatly improves the efficiency of crack monitoring and provides new data support for bridge structure health detection. It can be seen that the instantaneous concrete crack width change measurement scheme provided in this embodiment, by combining a high frame rate industrial camera and an improved deep learning network, realizes the accurate measurement of the instantaneous dynamic change of concrete crack width, effectively improves the accuracy of crack segmentation, and intuitively displays the change trend of crack width, thereby improving the efficiency and accuracy of bridge structure health monitoring.

[0159] It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0160] In addition, the present application also provides an electronic device, such as Fig.10 As shown, it shows a schematic diagram of the structure of the electronic device involved in the embodiment of the present application, specifically:

[0161] The electronic device may include components such as a processor 301 with one or more processing cores, a memory 302 with one or more computer-readable storage media, a power supply 303, and an input unit 304. Those skilled in the art will appreciate that Fig.10 The electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0162] The processor 301 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 302, and calling data stored in the memory 302, it executes various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 301 may include one or more processing cores; preferably, the processor 301 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 301.

[0163] The memory 302 can be used to store software programs and modules. The processor 301 executes various functional applications and the method for measuring instantaneous concrete crack width changes by running the software programs and modules stored in the memory 302. The memory 302 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the electronic device, etc. In addition, the memory 302 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 302 can also include a memory controller to provide the processor 301 with access to the memory 302.

[0164] The electronic device also includes a power supply 303 for supplying power to each component. Preferably, the power supply 303 can be logically connected to the processor 301 through a power management system, so that the power management system can manage charging, discharging, power consumption and other functions. The power supply 303 can also include one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, power status indicators and other arbitrary components.

[0165] The electronic device may further include an input unit 304, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0166] Although not shown, the electronic device may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 301 in the electronic device will load the executable files corresponding to the processes of one or more application programs into the memory 302 according to the following instructions, and the processor 301 will run the application programs stored in the memory 302, thereby realizing various functions, as follows:

[0167] Select the target concrete crack and paste the calibration object next to the target concrete crack; pre-set an industrial camera at a preset length from the target concrete, and use the industrial camera to collect the crack video corresponding to the target concrete crack; use the improved deep learning network to train the crack segmentation model to obtain the trained crack segmentation model; based on the trained crack segmentation model and the pre-integrated full-process algorithm, analyze the crack width change of the crack video to obtain the crack width change data corresponding to the target concrete crack and its visualization file.

[0168] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0169] The embodiments of the present application combine a high frame rate industrial camera and an improved deep learning network to achieve accurate measurement of the instantaneous dynamic changes in concrete crack width, effectively improve the accuracy of crack segmentation, and intuitively display the changing trend of crack width, thereby improving the efficiency and accuracy of bridge structure health monitoring.

[0170] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0171] To this end, an embodiment of the present application provides a storage medium in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute any of the steps in measuring instantaneous concrete crack width changes provided in the embodiment of the present application. For example, the instructions can execute the following steps:

[0172] Select the target concrete crack and paste the calibration object next to the target concrete crack; pre-set an industrial camera at a preset length from the target concrete, and use the industrial camera to collect the crack video corresponding to the target concrete crack; use the improved deep learning network to train the crack segmentation model to obtain the trained crack segmentation model; based on the trained crack segmentation model and the pre-integrated full-process algorithm, analyze the crack width change of the crack video to obtain the crack width change data corresponding to the target concrete crack and its visualization file.

[0173] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0174] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0175] Since the instructions stored in the storage medium can execute the steps in any method for measuring instantaneous concrete crack width changes provided in the embodiments of the present application, the beneficial effects that can be achieved by any method for measuring instantaneous concrete crack width changes provided in the embodiments of the present application can be achieved. For details, please refer to the previous embodiments and will not be repeated here.

[0176] The above is a detailed introduction to a method for measuring instantaneous concrete crack width changes provided in an embodiment of the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for measuring instantaneous concrete crack width change, characterized in that: The steps include: Select a target concrete crack and paste a calibration object next to the target concrete crack; An industrial camera is pre-installed at a preset length from the target concrete, and a crack video corresponding to the crack of the target concrete is collected by the industrial camera; An improved deep learning network is used to train the crack segmentation model to obtain a trained crack segmentation model; Based on the trained crack segmentation model and the pre-integrated full-process algorithm, crack width change analysis is performed on the crack video to obtain crack width change data corresponding to the target concrete crack and its visualization file.

2. The method for measuring instantaneous concrete crack width change according to claim 1, characterized in that: The method of pre-setting up an industrial camera at a preset length from the target concrete and collecting a crack video corresponding to the crack of the target concrete by the industrial camera includes: Setting up a camera bracket at a preset length from the target concrete, placing a working camera on the camera bracket, and connecting the working camera to a computer; The target concrete crack with the calibration object attached thereto is photographed by the working camera to obtain a calibrated concrete crack image; The calibration object is removed, and the crack video corresponding to the target concrete crack is collected by the working camera.

3. The method for measuring instantaneous concrete crack width change according to claim 1, characterized in that: The improved deep learning network is used to train the crack segmentation model to obtain a trained crack segmentation model, including: Select several crack images and their corresponding labels to construct a segmented image dataset, and preprocess the segmented image dataset; Using an improved deep learning network as the crack segmentation model network; The crack segmentation model network is trained based on the preprocessed segmented image data set to obtain a trained crack segmentation model.

4. The method for measuring instantaneous concrete crack width change according to claim 3, characterized in that: The improved deep learning network is used as the crack segmentation model network, comprising: Use the residual network structure ResNet-50 as the backbone feature extraction network; Adding an SE-Net module during the skip connection process of the backbone feature extraction network so that the backbone feature extraction network determines the importance of different channels by learning to assign different weights; An SK-Attention module is added to the upsampling of the network to enable the network to perform multi-scale feature fusion.

5. The method for measuring instantaneous concrete crack width change according to claim 1, characterized in that: The integration method of the full-process algorithm includes: Construct crack segmentation prediction algorithm, crack width calculation algorithm and result visualization algorithm; According to the crack analysis process, the crack segmentation prediction algorithm, the crack width calculation algorithm and the result visualization algorithm are integrated into a corresponding full-process algorithm.

6. The method for measuring instantaneous concrete crack width change according to claim 5, characterized in that: The crack width change analysis of the crack video based on the trained crack segmentation model and the pre-integrated full-process algorithm is performed to obtain the crack width change data corresponding to the target concrete crack and its visualization file, including: The crack video is segmented by using the crack segmentation prediction algorithm in the full-process algorithm to obtain a plurality of frames of crack segmentation images; The crack widths of the plurality of crack segmentation images are calculated respectively by the crack width calculation algorithm in the full-process algorithm to obtain corresponding crack width calculation results; The crack width calculation result is visualized by the result visualization algorithm in the full-process algorithm to obtain crack width change data corresponding to the target concrete crack and its visualization file.

7. The method for measuring instantaneous concrete crack width change according to claim 6, characterized in that: The crack video is segmented by the crack segmentation prediction algorithm in the full-process algorithm to obtain a plurality of frames of crack segmentation images, including: Performing frame-by-frame segmentation processing on the crack video to obtain a plurality of frames of crack segmentation images to be predicted; The crack segmentation prediction processing is performed on the plurality of frames of crack segmentation images to be predicted using a pre-trained crack segmentation model to obtain a plurality of corresponding crack segmentation images.

8. The method for measuring instantaneous concrete crack width change according to claim 6, characterized in that: The crack width calculation algorithm in the full-process algorithm is used to calculate the crack widths of the plurality of crack segmentation images to obtain corresponding crack width calculation results, including: Performing skeleton extraction on the plurality of crack segmentation image frames to obtain a central skeleton line corresponding to the crack in each crack segmentation image frame; An eight-direction calculation algorithm is used to search with the central skeleton line as the center to determine whether there are target pixel points in the crack outline in each direction and the number thereof; The target pixel points in the diagonal direction are geometrically merged, and the minimum pixel length is selected as the crack pixel width to obtain the crack width calculation result corresponding to each frame of the crack segmentation image.

9. The method for measuring instantaneous concrete crack width change according to claim 6, characterized in that: The method of visualizing the crack width calculation result by the result visualization algorithm in the full-process algorithm to obtain crack width change data corresponding to the target concrete crack and its visualization file includes: Performing data analysis on the crack width calculation results by using a result visualization algorithm to obtain a data file of the maximum width and average width corresponding to each frame of the crack segmentation image; Based on the data file of the maximum width and the average width corresponding to each frame of the crack segmentation image, a curve graph and a visual video showing the variation of the maximum width and the average width of the crack with the video time are generated.

10. The method for measuring instantaneous concrete crack width change according to claim 9, characterized in that: The method for generating the visualization video includes: Converting the color mode of the crack segmentation image from the RGB mode to the HSV mode; Determine the maximum crack width based on the data file of the maximum width and the average width corresponding to each frame of the crack segmentation image; Calculating the color width of each skeleton point based on the maximum crack width; Calculate the hue corresponding to each skeleton point based on the color width; According to the hue, preset saturation and brightness corresponding to each skeleton point, the color mode of the crack segmentation image is converted from the HSV mode to the RGB mode to generate a corresponding visualization video.

Images