Bridge crack detection method and system based on unmanned aerial vehicle

Abstract

The invention discloses a bridge crack detection method and system based on an unmanned aerial vehicle, relates to the technical field of image recognition, and solves the problems that in the prior art, when the crack width is measured, depending on the connectivity of binary image segmentation, if holes or fractures exist in the binary segmentation process, the result of intercept calculation has errors, and the detection accuracy is poor. And deviation exists between the identified crack and the actual crack. The method comprises the following steps: acquiring real-time images of a plurality of bridges through an unmanned aerial vehicle, and preprocessing the real-time images to obtain bridge images; synthesizing a bridge panorama by using an SLAM technology according to the plurality of bridge images; recognizing a bridge crack according to the bridge panorama to obtain a crack image; analyzing the crack according to the crack image to obtain crack data, and analyzing the risk level of the crack based on the crack data; processing the bridge crack according to the risk level of the crack; the bridge crack can be accurately identified, the bridge crack can be processed in time, and the use safety of the bridge can be guaranteed.

Description

Technical Field

[0001] The present invention belongs to the field of image recognition, relates to bridge crack detection technology, and specifically is a method and system for bridge crack detection based on unmanned aerial vehicles (UAVs). Background Art

[0002] Bridge cracks have an important impact on the safety and service life of bridges; cracks will reduce the load-bearing capacity of bridges and affect the normal use functions of bridges; by regularly detecting cracks, potential structural problems can be discovered in time to prevent serious safety accidents such as bridge collapses caused by crack expansion. Crack detection helps to understand the current state of the bridge structure, so as to take appropriate maintenance and reinforcement measures; through crack detection, the damage degree and speed of each part of the bridge can be evaluated, so as to formulate a more scientific and reasonable maintenance plan; it helps to avoid unnecessary maintenance activities, reduce maintenance costs and improve maintenance efficiency; bridge cracks may cause bumps or instability when vehicles are driving, increasing the risk of traffic accidents; detecting and repairing cracks can reduce this risk and improve the safety of road users.

[0003] The prior art (a patent application for an invention patent with a publication number of CN117911371A) discloses a method for intelligent identification and measurement of bridge cracks. The method includes: Step 1: Input the bridge crack data set into the YOLOv8 network for transfer learning to obtain a target recognition and positioning model. Input the bridge image to be processed into the target recognition and positioning model to identify and position the cracks in the image, and output the bridge crack image; Step 2: Establish an ultra-high-definition image segmentation model, and input the bridge crack image output in Step 1 into the ultra-high-definition image segmentation model to obtain a binary image of the bridge cracks; Step 3: Measure the pixel width of the binary image of the cracks obtained in Step 2 based on a hybrid method of the shortest distance method and the orthogonal skeleton method. The prior art uses a model to identify the cracks in the bridge image. However, when measuring the crack width, it depends on the connectivity of binary image segmentation. If there are holes or breaks in the binary segmentation process, it will lead to a large error in the result of intercept calculation, causing a deviation between the identified cracks and the actual situation.

[0004] The present invention provides a method and system for bridge crack detection based on UAVs to solve the above technical problems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; for this purpose, the present invention proposes a method and system for bridge crack detection based on UAVs, which is used to solve the technical problem that the prior art uses a model to identify the cracks in the bridge image. However, when measuring the crack width, it depends on the connectivity of binary image segmentation. If there are holes or breaks in the binary segmentation process, it will lead to a large error in the result of intercept calculation, causing a deviation between the identified cracks and the actual situation.

[0006] To achieve the above object, a first aspect of the present invention provides a method for detecting bridge cracks based on an unmanned aerial vehicle, including:

[0007] Step S1: Collect real-time images of several bridges through an unmanned aerial vehicle, and preprocess the real-time images to obtain bridge images;

[0008] Step S2: Synthesize a panoramic view of the bridge using SLAM technology based on several bridge images; identify bridge cracks based on the panoramic view of the bridge to obtain crack images;

[0009] Step S3: Analyze the cracks based on the crack images to obtain crack data, and analyze the risk level of the cracks based on the crack data;

[0010] Step S4: Process the bridge cracks according to the risk level of the cracks.

[0011] Preferably, the preprocessing of the real-time images to obtain bridge images includes:

[0012] Retrieve several real-time images; analyze the clarity of the real-time images according to a clarity threshold; wherein, the clarity threshold includes a primary threshold and a secondary threshold;

[0013] When the clarity is less than the primary threshold, the corresponding image is recollected; otherwise, the clarity is compared with the secondary threshold;

[0014] When the clarity is less than the secondary threshold, the real-time images are processed using Gaussian filtering; the adjusted and recollected real-time images are cropped and brightness-adjusted, and the processed real-time images are marked as bridge images.

[0015] It should be noted that the primary threshold and the secondary threshold are set by expert evaluation; and the primary threshold is less than the secondary threshold;

[0016] The present invention analyzes the clarity of several real-time images. When the clarity does not meet the standard, the clarity of the real-time images is adjusted or the corresponding real-time images are recollected, and the size and brightness of the real-time images are adjusted, which can ensure the image quality of the real-time images and provide a basis for subsequent crack detection.

[0017] Preferably, the generation of the panoramic view of the bridge using SLAM technology based on several bridge images includes:

[0018] Retrieve several bridge images, extract key feature point clouds from the several bridge images using a feature point algorithm; associate the key feature point clouds in adjacent bridge images using a feature point descriptor;

[0019] Analyze the associated key feature point cloud using visual odometry to obtain the motion parameters of the UAV in three-dimensional space; calculate the three-dimensional coordinates of the key feature point cloud according to the motion parameters, and construct an initial three-dimensional sparse point cloud image;

[0020] Use multi-view image restoration technology to restore the three-dimensional sparse point cloud image to obtain a dense point cloud image; project several bridge images onto the surface of the dense point cloud image to obtain a panoramic bridge image.

[0021] It should be noted that the feature point algorithm is an algorithm used to extract features with obvious changes or unique interests from images, including the SIFT algorithm or the ORB algorithm; the feature point descriptor is a method for numerically describing the extracted feature points, which converts the local information around the feature points into a mathematical description in the form of a vector or matrix, including: SIFT descriptor or ORB descriptor; visual odometry is a process of determining the position and orientation of a robot by analyzing relevant camera images, using continuous camera images to estimate the travel distance, so as to determine the equivalent odometry information; multi-view restoration technology usually refers to multi-view image reconstruction technology, that is, the process of estimating the corresponding three-dimensional structure from a sequence of two-dimensional images that may combine local motion signals, aiming to enable the computer to have a function similar to the human visual system and be able to reconstruct the three-dimensional structure of an object through the two-dimensional image information captured.

[0022] Preferably, the identification of bridge cracks based on the panoramic bridge image to obtain a crack image includes:

[0023] Retrieve the panoramic bridge image, and convert the panoramic bridge image into a bridge floor plan by projection; divide the bridge floor plan into several identification images and number them;

[0024] Use an edge detection algorithm to identify the contours of several identification images to obtain a detection image; integrate several detection images according to the numbers to obtain a detection input sequence;

[0025] Retrieve the crack detection model, and input the detection input sequence into the crack detection model to obtain the corresponding crack image; among them, the crack detection model is constructed based on an artificial intelligence model.

[0026] The present invention converts the panoramic bridge image into a bridge floor plan, divides the bridge floor plan into several identification images; performs edge detection on several identification images to obtain a detection image; uses a crack detection model to identify the crack image; can identify the cracks of the bridge, improve the accuracy of identification, and the partition identification is beneficial to improving the identification efficiency.

[0027] Preferably, the use of an edge detection algorithm to identify the contours of several identification images to obtain a detection image includes:

[0028] Retrieve a number of recognition images, and perform grayscale processing on the number of recognition images using a grayscale function; process the number of recognition images using a Gaussian filter smoother to obtain a Gaussian kernel convolution;

[0029] Set up a pair of convolutional arrays and where G x represents the convolutional array in the x-axis direction; G y represents the convolutional array in the y-axis direction;

[0030] Calculate the gradient magnitude through the formula Calculate the gradient direction through the formula Compare the magnitudes of adjacent pixels according to the gradient direction. When the magnitude is not the maximum, mark the corresponding magnitude as 0;

[0031] Connect the edges of the image according to the set double thresholds to obtain the contour of the recognition image; mark the recognition image after edge detection as the detection image; where the double thresholds include: a low threshold and a high threshold.

[0032] It should be noted that the double thresholds are set according to simulation experiments. The low threshold is used to detect weak edges, and the high threshold is used to detect strong edges; usually, high threshold: low threshold = 3:1 or 2:1; using double thresholds to detect edges can connect discontinuous edges.

[0033] Preferably, the crack detection model is constructed based on an artificial intelligence model, including:

[0034] Obtain a standard image set; where the standard image set includes: standard input images consistent with the content attributes of the detection input sequence, and standard input images consistent with the content attributes of the crack images;

[0035] Divide the standard image set into a training set, a validation set, and a test set according to a set ratio; use the training set to train the artificial intelligence model; use the validation set to adjust the internal parameters in the artificial intelligence model; use the test set to test the artificial intelligence model to obtain test metrics;

[0036] Judge whether the test metrics are greater than the metric threshold; if yes, mark the artificial intelligence model as the crack detection model; if not, retrain the artificial intelligence model.

[0037] It should be noted that according to a number of bridge cracks and bridge panoramic images for processing, a standard image set is integrated; the ratio of the division of the standard image set is set after being evaluated by experts; the test metrics include: accuracy, F1 score, recall rate, and stability; the metric threshold is set through experimental simulation.

[0038] Preferably, analyzing the cracks based on the crack images to obtain crack data includes:

[0039] Retrieve the bridge plan view to obtain the actual size of the bridge; measure the bridge size in the bridge plan view, compare the bridge size with the actual size to obtain the scaling ratio;

[0040] Retrieve the crack images, identify the number of cracks from the crack images; measure the length and width of the cracks respectively; calculate the length and width of the cracks according to the scaling ratio to obtain the crack length and crack width; integrate the number of cracks, crack length, and crack width into crack data.

[0041] The present invention obtains the corresponding scaling ratio according to the bridge plan view and the actual size of the bridge; obtains the crack size according to the scaling ratio and the size in the crack images, and integrates the number of cracks and the crack size into crack data; provides data support for subsequent analysis of the risk level of cracks; is conducive to evaluating the risk of bridge cracks in combination with the actual situation.

[0042] Preferably, analyzing the risk level of the cracks based on the crack data includes:

[0043] Retrieve the crack data; number the crack length and crack width in the crack data according to the number of cracks and mark them as LCj and LKj respectively; mark the number of cracks in the crack data as LG;

[0044] Through the formula Calculate the risk coefficient of the bridge cracks; where α and β are proportionality coefficients greater than 0, and DL is the dimension removal coefficient; j represents the crack number, j = 1, 2,..., m, and m is a positive integer;

[0045] Compare the risk coefficient of the bridge cracks with the risk threshold respectively to obtain the corresponding interval of the risk coefficient; divide the risk level of the cracks according to the interval of the risk coefficient.

[0046] The present invention calculates the risk coefficient of the cracks according to the crack data; divides the risk level of the cracks according to the risk coefficient of the cracks; can evaluate the risk of the cracks, and timely process the bridges with high risks, which is conducive to ensuring the safety of bridge use.

[0047] Preferably, treating the bridge cracks according to the risk level of the cracks includes:

[0048] Retrieve the risk level of the cracks to obtain the risk treatment library; match the risk level with the risk treatment library to obtain the corresponding crack treatment measures;

[0049] Send the corresponding crack treatment measures to the bridge management personnel, and dispatch technical personnel with the corresponding number according to the crack treatment measures to treat the bridge cracks.

[0050] It should be noted that when multiple bridges need to be treated for bridge cracks, the bridges are sorted according to the risk level; the bridges with high risk are treated first, and then the bridges with low risk are treated.

[0051] The present invention matches corresponding crack treatment measures according to the risk level of cracks, and dispatches technical personnel with corresponding numbers according to the crack treatment measures to treat bridge cracks, which can timely treat bridges with risks, is beneficial to ensuring the safety of bridge drivers, and prolongs the service life of bridges.

[0052] The second aspect of the present invention provides a bridge crack detection system based on an unmanned aerial vehicle, including: a crack detection module, a risk processing module connected thereto, and an image acquisition module;

[0053] Image acquisition module: used to collect real-time images of several bridges through an unmanned aerial vehicle, and preprocess the real-time images to obtain bridge images;

[0054] Crack detection module: used to synthesize a panoramic view of the bridge using SLAM technology according to several bridge images; identify bridge cracks according to the panoramic view of the bridge to obtain crack images; analyze the cracks according to the crack images to obtain crack data;

[0055] Risk processing module: used to analyze the risk level of cracks based on crack data; treat bridge cracks according to the risk level of cracks.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] 1. The present invention analyzes the clarity of several real-time images. When the clarity does not meet the standard, the clarity of the real-time images is adjusted or the corresponding real-time images are re-collected, and the size and brightness of the real-time images are adjusted, which can ensure the image quality of the real-time images and provide a basis for subsequent crack detection; the panoramic view of the bridge is converted into a plane view of the bridge, and the plane view of the bridge is divided into several recognition images; edge detection is performed on the several recognition images to obtain detection images; a crack detection model is used to identify the crack images; the cracks of the bridge can be identified, the accuracy of the identification is provided, and the partition identification is beneficial to improving the identification efficiency.

[0058] 2. The present invention obtains the corresponding scaling ratio according to the bridge plan and the actual size of the bridge; obtains the crack size based on the scaling ratio and the size in the crack image, and integrates the number of cracks and the crack size into crack data, providing data support for subsequent analysis of the risk level of cracks; facilitating the assessment of the risk of bridge cracks in combination with the actual situation; calculating the risk coefficient of cracks according to the crack data; dividing the risk level of cracks according to the risk coefficient of cracks; being able to assess the risk of cracks and promptly handle bridges with high risks, which is beneficial to ensuring the safety of bridge use; matching corresponding crack treatment measures according to the risk level of cracks, and dispatching technicians with corresponding numbers according to the crack treatment measures to handle the bridge cracks, being able to promptly handle bridges with existing risks, which is beneficial to ensuring the safety of bridge drivers and extending the service life of bridges. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0060] Figure 1 Schematic diagram of the overall steps of the system of the present invention;

[0061] Figure 2 Schematic diagram of the crack image recognition steps of the present invention;

[0062] Figure 3 Schematic diagram of the crack risk assessment steps of the present invention;

[0063] Figure 4 Schematic diagram of the specific steps of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0064] The following will clearly and completely describe the technical solutions of the present invention in combination with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0065] Please refer to Figure 1 , the first aspect embodiment of the present invention provides a method for detecting bridge cracks based on an unmanned aerial vehicle, including:

[0066] Step S1: Collect real-time images of several bridges through an unmanned aerial vehicle, and preprocess the real-time images to obtain bridge images;

[0067] Step S2: Synthesize a panoramic bridge image using SLAM technology based on several bridge images; identify bridge cracks based on the panoramic bridge image to obtain crack images;

[0068] Step S3: Analyze the cracks based on the crack images to obtain crack data, and analyze the risk levels of the cracks based on the crack data;

[0069] Step S4: Process the bridge cracks according to the risk levels of the cracks.

[0070] Please refer to Figure 2 , real-time images of several bridges are collected by a drone, and the clarity of the real-time images is analyzed according to a clarity threshold; wherein, the clarity threshold includes a primary threshold and a secondary threshold; when the clarity is less than the primary threshold, the corresponding image is collected again; otherwise, the clarity is compared with the secondary threshold; when the clarity is less than the secondary threshold, the real-time image is processed using Gaussian filtering; the adjusted and re-collected real-time images are cropped and brightness-adjusted, and the processed real-time images are marked as bridge images;

[0071] Extract key feature point clouds from several bridge images using a feature point algorithm; associate the key feature point clouds in adjacent bridge images using a feature point descriptor; analyze the associated key feature point clouds using a visual odometer to obtain the motion parameters of the drone in three-dimensional space; calculate the three-dimensional coordinates of the key feature point clouds according to the motion parameters, and construct an initial three-dimensional sparse point cloud image; restore the three-dimensional sparse point cloud image using a multi-view image restoration technology to obtain a dense point cloud image; project several bridge images onto the surface of the dense point cloud image to obtain a panoramic bridge image.

[0072] Convert the panoramic bridge image into a bridge plan using a projection method; divide the bridge plan into several recognition images and number them; identify the contours of the several recognition images using an edge detection algorithm to obtain detection images; integrate the several detection images according to the numbers to obtain a detection input sequence; retrieve a crack detection model, and input the detection input sequence into the crack detection model to obtain corresponding crack images; wherein, the crack detection model is constructed based on an artificial intelligence model.

[0073] It should be noted that identifying the contours of several recognition images using an edge detection algorithm to obtain detection images includes:

[0074] Retrieve several recognition images, and perform grayscale processing on the several recognition images using a grayscale function; process the several recognition images using a Gaussian filter smoother to obtain a Gaussian kernel convolution;

[0075] Set a pair of convolutional arrays and wherein, G xThe convolution array in the x-axis direction; G y The convolution array in the y-axis direction;

[0076] Calculate the gradient magnitude through the formula Calculate the gradient direction through the formula Compare the magnitudes of adjacent pixels according to the gradient direction. When the magnitude is not the maximum, mark the corresponding magnitude as 0;

[0077] Connect the edges of the image according to the set double thresholds to obtain the contour of the recognized image; Mark the recognized image after edge detection as the detected image; Among them, the double thresholds include: a low threshold and a high threshold.

[0078] It should be noted that the double thresholds are set according to simulation experiments. The low threshold is used to detect weak edges, and the high threshold is used to detect strong edges; Usually, high threshold: low threshold = 3:1 or 2:1; Using the double thresholds to detect edges can connect discontinuous edges.

[0079] It is worth noting that the crack detection model is constructed based on an artificial intelligence model, including:

[0080] Obtain a standard image set; Among them, the standard image set includes: standard input images consistent with the content attributes of the detected input sequence, and standard input images consistent with the content attributes of the crack images;

[0081] Divide the standard image set into a training set, a validation set, and a test set according to a set ratio; Use the training set to train the artificial intelligence model; Use the validation set to adjust the internal parameters in the artificial intelligence model; Use the test set to test the artificial intelligence model to obtain test metrics;

[0082] Judge whether the test metric is greater than the metric threshold; If yes, mark the artificial intelligence model as the crack detection model; If not, retrain the artificial intelligence model.

[0083] It should be noted that according to a number of bridge cracks and bridge panoramic images for processing, a standard image set is integrated; The division ratio of the standard image set is set after being evaluated by experts; The test metrics include: accuracy, F1 score, recall rate, and stability; The metric threshold is set through experimental simulation.

[0084] It should be noted that preprocessing the collected bridge images can ensure the image quality of the bridge images, facilitate subsequent analysis of the bridge images; Construct a panoramic image of the bridge image, convert the panoramic image into a plan view, divide the bridge plan view into several recognized images, perform edge detection on the several recognized images to obtain the detected image, and use the model to recognize the detected image to obtain the crack image; It can improve the recognition accuracy and is conducive to laying a data foundation for subsequent analysis of the cracks.

[0085] Please refer to Figure 3 to retrieve the bridge plan view, obtain the actual dimensions of the bridge; measure the bridge dimensions in the bridge plan view, compare the bridge dimensions with the actual dimensions to obtain the scaling ratio; retrieve the crack image, identify the number of cracks from the crack image; measure the length and width of the cracks respectively; calculate the length and width of the cracks according to the scaling ratio to obtain the crack length and crack width; integrate the number of cracks, crack length and crack width into crack data;

[0086] Number and mark the crack length and crack width in the crack data according to the number of cracks as LCj and LKj respectively; mark the number of cracks in the crack data as LG; through the formula calculate the risk coefficient of the bridge cracks; where α and β are proportionality coefficients greater than 0, and DL is the dimension removal coefficient; j represents the crack number, j = 1, 2,..., m, and m is a positive integer; compare the risk coefficients of the bridge cracks with the risk thresholds respectively to obtain the corresponding risk coefficient intervals; divide the risk levels of the cracks according to the risk coefficient intervals.

[0087] It should be noted that the proportionality coefficients are set by expert evaluation, and the dimension removal coefficient DL is used to remove the dimensions in the formula, generally set to 1; when the number of cracks in the bridge is larger, the corresponding value of α×LG is larger, so the risk coefficient is larger; when the crack length and width are larger, the corresponding value is larger; so the risk coefficient is larger; in summary, the risk coefficient is positively correlated with the crack data.

[0088] Retrieve the risk treatment library; match the risk level with the risk treatment library to obtain the corresponding crack treatment measures; send the corresponding crack treatment measures to the bridge management personnel, and dispatch technical personnel with the corresponding number according to the crack treatment measures to deal with the bridge cracks.

[0089] It should be noted that when there are multiple bridges that need to deal with the bridge cracks, the bridges are sorted according to the risk level; first deal with the bridges with high risk, and then deal with the bridges with low risk.

[0090] It should be noted that analyzing the crack data based on the crack image, calculating the risk coefficient of the cracks according to the crack data, comparing the risk coefficient with the risk threshold to obtain the risk level of the cracks; determining the crack treatment measures based on the risk level, and dispatching technical personnel with the corresponding number according to the treatment measures to deal with the cracks; can timely deal with the bridge cracks, which is beneficial to ensuring the safety of bridge use.

[0091] In the second aspect of the embodiments of the present invention, a bridge crack detection system based on an unmanned aerial vehicle (UAV) is provided, including: a crack detection module, a risk processing module connected thereto, and an image acquisition module;

[0092] Image acquisition module: It is used to collect real-time images of several bridges through a UAV and preprocess the real-time images to obtain bridge images;

[0093] Crack detection module: It is used to synthesize a panoramic view of the bridge using SLAM technology based on several bridge images; identify bridge cracks according to the panoramic view of the bridge to obtain crack images; analyze the cracks based on the crack images to obtain crack data;

[0094] Risk processing module: It is used to analyze the risk level of the cracks based on the crack data; process the bridge cracks according to the risk level of the cracks.

[0095] Some of the data in the above formula are taken as numerical values after removing the dimension. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation; the preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0096] Working principle of the present invention: The present invention collects real-time images of several bridges through a UAV, preprocesses the real-time images to obtain bridge images; generates a panoramic view of the bridge using SLAM technology based on several bridge images; identifies bridge cracks according to the panoramic view of the bridge to obtain crack images; analyzes the cracks based on the crack images to obtain crack data, analyzes the risk level of the cracks based on the crack data; processes the bridge cracks according to the risk level of the cracks.

[0097] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for detecting bridge cracks based on an unmanned aerial vehicle, characterized in that, Including: Step S1: Collect real-time images of several bridges by using a drone, and preprocess the real-time images to obtain bridge images; Step S2: Synthesize a panoramic bridge image by using the SLAM technology based on several bridge images; Identify bridge cracks based on the panoramic bridge image to obtain crack images; Step S3: Analyze the cracks based on the crack images to obtain crack data, and analyze the risk level of the cracks based on the crack data; Step S4: Process the bridge cracks according to the risk level of the cracks.

2. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 1, wherein, The preprocessing of the real-time images to obtain bridge images includes: Retrieve several real-time images; Analyze the clarity of the real-time images according to a clarity threshold; Among them, the clarity threshold includes a primary threshold and a secondary threshold; When the clarity is less than the primary threshold, re-collect the corresponding image; Otherwise, compare the clarity with the secondary threshold; When the clarity is less than the secondary threshold, process the real-time images by using Gaussian filtering; Crop and adjust the brightness of the adjusted and re-collected real-time images, and mark the processed real-time images as bridge images.

3. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 1, wherein The generating of the panoramic bridge image by using the SLAM technology based on several bridge images includes: Retrieve several bridge images, and extract key feature point clouds from the several bridge images by using a feature point algorithm; Associate the key feature point clouds in adjacent bridge images by using a feature point descriptor; Analyze the associated key feature point clouds by using a visual odometer to obtain the motion parameters of the drone in three-dimensional space; Calculate the three-dimensional coordinates of the key feature point clouds according to the motion parameters, and construct an initial three-dimensional sparse point cloud image; Restore the three-dimensional sparse point cloud image by using a multi-view image restoration technology to obtain a dense point cloud image; Project the several bridge images onto the surface of the dense point cloud image to obtain a panoramic bridge image.

4. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 1, wherein, The identifying of the bridge cracks based on the panoramic bridge image to obtain crack images includes: Retrieve the panoramic bridge image, and convert the panoramic bridge image into a bridge plan view by using a projection method; Divide the bridge plan view into several identification images and number them; Identify the contours of the several identification images by using an edge detection algorithm to obtain detection images; Integrate the several detection images according to the numbers to obtain a detection input sequence; Retrieve a crack detection model, and input the detection input sequence into the crack detection model to obtain corresponding crack images; Among them, the crack detection model is constructed based on an artificial intelligence model.

5. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 4, characterized in that, The identifying of the contours of the several identification images by using an edge detection algorithm to obtain detection images includes: Retrieve the several identification images, and perform grayscale processing on the several identification images by using a grayscale function; Process the several identification images by using a Gaussian filter smoother to obtain a Gaussian kernel convolution; Set a pair of convolutional arrays and where G x represents the convolutional array in the x-axis direction; G y represents the convolutional array in the y-axis direction; Calculate the gradient magnitude through the formula Calculate the gradient direction through the formula Compare the magnitudes of adjacent pixels according to the gradient direction. When the magnitude is not the maximum, mark the corresponding magnitude as 0; Connect the edges of the image according to set double thresholds to obtain the contours of the identification images; Mark the identification images after edge detection as detection images; Among them, the double thresholds include: a low threshold and a high threshold.

6. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 4, wherein The crack detection model is constructed based on an artificial intelligence model, including: Obtain a standard image set; Among them, the standard image set includes: standard input images consistent with the content attributes of the detection input sequence, and standard input images consistent with the content attributes of the crack images; Divide the standard image set into a training set, a validation set, and a test set according to a set ratio; use the training set to train the artificial intelligence model; use the validation set to adjust the internal parameters in the artificial intelligence model; use the test set to test the artificial intelligence model to obtain test metrics; Judge whether the test metrics are greater than the metric threshold; if so, mark the artificial intelligence model as a crack detection model; if not, retrain the artificial intelligence model.

7. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 1, characterized in that The analysis of the cracks based on the crack images to obtain crack data includes: Retrieve the bridge floor plan and obtain the actual dimensions of the bridge; measure the bridge dimensions in the bridge floor plan and compare the bridge dimensions with the actual dimensions to obtain the scaling ratio; Retrieve the crack images, identify the number of cracks from the crack images; measure the length and width of the cracks respectively; calculate the length and width of the cracks according to the scaling ratio to obtain the crack length and crack width; integrate the number of cracks, the crack length, and the crack width into crack data.

8. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 1, wherein, The analysis of the risk level of the cracks based on the crack data includes: Retrieve the crack data; number the crack length and crack width in the crack data according to the number of cracks and mark them as LCj and LKj respectively; mark the number of cracks in the crack data as LG; Calculate the risk coefficient of bridge cracks through the formula ; where α and β are proportionality coefficients greater than 0, DL is the dimensionless removal coefficient; j represents the crack number, j = 1, 2, …, m, and m is a positive integer Compare the risk coefficients of the bridge cracks with the risk thresholds respectively to obtain the corresponding intervals of the risk coefficients; divide the risk levels of the cracks according to the intervals of the risk coefficients.

9. The method for detecting bridge cracks based on an unmanned aerial vehicle according to claim 1, wherein The treatment of the bridge cracks according to the risk levels of the cracks includes: Retrieve the risk levels of the cracks and obtain the risk treatment library; match the risk levels with the risk treatment library to obtain the corresponding crack treatment measures; Send the corresponding crack treatment measures to the bridge management personnel, and dispatch technical personnel with the corresponding number according to the crack treatment measures to treat the bridge cracks.

10. The drone-based bridge crack detection system is applied to the drone-based bridge crack detection method according to any one of claims 1-9, and is characterized in that, It includes: A crack detection module, and a risk treatment module and an image acquisition module connected thereto; The image acquisition module: used to collect real-time images of several bridges through a drone, and preprocess the real-time images to obtain bridge images; The crack detection module: used to synthesize a bridge panoramic view according to several bridge images by using the SLAM technology; identify bridge cracks according to the bridge panoramic view to obtain crack images; analyze the cracks based on the crack images to obtain crack data; The risk treatment module: used to analyze the risk level of the cracks based on the crack data; Treat the bridge cracks according to the risk levels of the cracks.

Images