Bridge crack image detection and health monitoring system based on unmanned aerial vehicle

Abstract

The invention discloses a bridge crack image detection and health monitoring system based on an unmanned aerial vehicle, and relates to the technical field of bridge health monitoring, the bridge crack image detection and health monitoring system comprises an unmanned aerial vehicle module, a computing platform and a database, the database transmits a monitoring model of a corresponding bridge to the computer platform; the computing platform sets a plurality of position matching points based on fixed parts of the bridge in the monitoring model, and the monitoring model is divided into a plurality of sub-models based on the position matching points; each sub-model is provided with at least one group of feature points; and the unmanned aerial vehicle module plans a flight path based on each position matching point in the monitoring model, acquires detection image data of the corresponding bridge when the unmanned aerial vehicle module moves along the flight path, and transmits the detection image data to the computing platform. According to the method, the overall bridge model is decomposed into a plurality of sub-models, the complex structure of the bridge is decomposed into a plurality of independent units, and accurate positioning and detection of key parts are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge health monitoring, and specifically provides a bridge crack image detection and health monitoring system based on an unmanned aerial vehicle (UAV). Background Art

[0002] With the rapid development of infrastructure, the number and scale of bridges have been increasing year by year. As an important transportation facility, the safety of bridges is directly related to the safety of public travel. However, during the long-term use of bridges, due to the influence of factors such as loads and environmental corrosion, structural problems such as cracks and deformations often occur. If these problems are not discovered and repaired in time, serious safety accidents may be caused.

[0003] After retrieval, Chinese Patent (Publication No.: CN118134871A) discloses a method and system for quickly obtaining actual crack information based on an unmanned aerial vehicle. The patent includes: preprocessing a bridge structure image, obtaining a crack binary image based on the preprocessed bridge structure image, a crack recognition model, and a crack segmentation model; obtaining the pixel area of the crack based on the crack binary image; extracting the crack contour based on the crack binary image, performing polynomial fitting on the crack contour to obtain a fitting curve, obtaining the optimal fitting curve based on the root mean square error of the fitting curve, and obtaining the pixel length of the crack based on the optimal fitting curve and the crack contour; obtaining the pixel width of the crack based on the two intersection points of the normal line of the fitting curve and the crack contour; and obtaining the actual geometric dimension information of the crack based on the similarity principle according to the pixel area, pixel length, and pixel width of the crack.

[0004] In the prior art, a single detection method has poor adaptability in complex environments, is prone to false detection or missed detection, and most of the existing methods are static detections, which cannot effectively monitor the expansion trend of cracks and are difficult to comprehensively evaluate the dynamic health status of bridges. Therefore, the present invention proposes a bridge crack image detection and health monitoring system based on an unmanned aerial vehicle. Summary of the Invention

[0005] The purpose of the present invention is to provide a bridge crack image detection and health monitoring system based on an unmanned aerial vehicle to solve the problems mentioned in the above background art.

[0006] The present invention can be realized through the following technical solutions: A bridge crack image detection and health monitoring system based on an unmanned aerial vehicle, including a UAV module, a computing platform, and a database;

[0007] The database performs three-dimensional modeling on the bridge to obtain a monitoring model of the corresponding bridge;

[0008] The computing platform is used to receive a monitoring model. In the monitoring model, multiple position matching points are set based on the fixed parts of the bridge. The monitoring model is divided into multiple sub-models based on each position matching point, and at least one set of feature points is set for each sub-model.

[0009] The position matching points are key parts of the bridge such as support points, connection points, expansion joints, etc. The feature points represent information such as the geometric shape and structural characteristics of the area. Through the geometric and texture information of the 3D model, the extracted feature points will be used as the "template" feature points of the sub-model.

[0010] The UAV module plans a flight path based on each position matching point in the monitoring model. When the UAV module moves along the flight path, it acquires the detection image data of the corresponding bridge and transmits the detection image data to the computing platform.

[0011] The computing platform intercepts the detection image data within the corresponding time period based on the matching time between the UAV module and each position matching point to obtain the node image data, and the computing platform matches each node image data with the sub-model corresponding to the position matching point in the monitoring model.

[0012] The computing platform establishes a data set for each sub-model to collect the node image data obtained by the UAV module each time, and the computing platform analyzes the time series of each data set to obtain the health data of the corresponding sub-model part of the bridge.

[0013] At the same time, the computing module transmits the node image data in each data set to the database for backup.

[0014] A further technical improvement of the present invention is that when the UAV module acquires the detection image data, it adds a timestamp to the detection image data.

[0015] The computing platform synchronizes the flight path of the UAV module with the timestamp of the position matching point based on the predetermined time period of the position matching point. The computing platform first obtains the time period corresponding to each position matching point to ensure the time matching between each position matching point and the node image data.

[0016] A further technical improvement of the present invention is that the method for the computing platform to match each node image data with the sub-model includes the following steps:

[0017] S1. The computer platform extracts the key points in the image from each node image data, and each key point is respectively matched with the feature points in the corresponding sub-model.

[0018] S2. The computer platform matches the key points extracted from the node image data with the feature points in the corresponding sub-model through a feature matching algorithm.

[0019] The matching algorithm calculates the similarity based on the descriptors of the feature points and selects the best matching points;

[0020] S3. After the feature matching is completed, the computer platform calculates the geometric transformation matrix based on the matched key points and feature points;

[0021] The computer platform corrects the image by using perspective transformation to align the feature points in the node image with the position matching points in the 3D monitoring model;

[0022] S4. According to the calculated geometric transformation matrix, the computer platform performs geometric transformation on the image to align the geometric shape of the image in the node image data with the geometric shape of the corresponding sub-model, ensuring the accurate spatial matching between the node image data and the sub-model.

[0023] A further technical improvement of the present invention lies in that: the computer platform divides the node image data in the sub-model dataset into grids to generate multiple recognition parts, and uses the gray-scale difference and time series analysis between the recognition parts to perform bridge crack detection and health monitoring. The specific steps are as follows:

[0024] Y1. The computer platform divides each node image data I into small areas of m×n according to the grid, and each small area is a recognition part;

[0025] Y2. Calculate the average gray-scale value of each recognition part to generate the corresponding gray-scale image;

[0026] Y3. The computer platform calculates the gray-scale difference between adjacent recognition parts in each node image data;

[0027] Y4. The computer platform performs time series analysis on the gray-scale difference between the same recognition parts at different time points t;

[0028] Y5. The computer platform evaluates the health status of the corresponding part of the bridge based on the gray-scale difference and time series change between each recognition part, specifically including:

[0029] a1. The computer platform sets a difference threshold to determine whether there is an abnormal gray-scale difference;

[0030] a2. Aggregate the time series change trend to analyze the crack propagation speed;

[0031] And the computer platform generates a health score according to the gray-scale difference and time series change of the recognition part.

[0032] A further technical improvement of the present invention lies in that: before the computer platform recognizes the gray-scale differences of each recognition part, histogram equalization is performed on the node image data of each recognition part. By redistributing the gray-scale values of the image pixels, the gray-scale distribution of the image becomes more uniform, enhancing the contrast.

[0033] A further technical improvement of the present invention lies in that: the shown unmanned aerial vehicle module includes an infrared unit for acquiring infrared images of the bridge. And when the infrared unit acquires infrared images of the bridge, the same time stamp as the detected image data is added to the infrared images to ensure that the spatial positions of the subsequent generated gray-scale images of the infrared images and the detected image data correspond one by one.

[0034] A further technical improvement of the present invention lies in that: the computer platform also divides the infrared image into multiple infrared recognition parts according to an m×n grid, calculates the average temperature of each infrared recognition part, and calculates the temperature difference between adjacent infrared recognition parts;

[0035] And the computer platform combines the temperature difference between adjacent infrared recognition parts with the gray-scale difference of the corresponding node image data to form a comprehensive feature difference;

[0036] The formula for the comprehensive feature difference is ΔF i,j,j+1 =β1G i,j,j+1 +β2ΔT i,j,j+1 ;

[0037] In the formula, ΔF i,j,j+1 is the comprehensive feature difference, β1 and β2 are weight parameters, representing the contribution ratios of the gray-scale difference and the temperature difference to the detection, which are determined through experiments;

[0038] G i,j,j+1 is the gray-scale difference of the corresponding node image data; ΔT i,j,j+1 is the temperature difference between adjacent infrared recognition parts;

[0039] And the computer platform generates a health score according to the time series change of the comprehensive feature difference.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The present invention proposes to decompose the overall bridge model into multiple sub-models, and establish recognition regions based on position-matching points for each sub-model. Through this method of sub-model partitioning, the complex bridge structure is decomposed into multiple independent units, achieving precise positioning and detection of key parts (such as support points, connection points, expansion joints, etc.). At the same time, independent data sets are established for each sub-model, which can be refined to specific parts of the bridge to evaluate the health status of different regions. By comprehensively analyzing the gray-scale difference, temperature difference characteristics, and dynamic change trends, a health score is generated for each sub-model, and finally the overall health status of the bridge is quantitatively evaluated;

[0042] Moreover, through time-series analysis of the data sets of the sub-models, the present invention can dynamically track the expansion trend of cracks. By analyzing the changes in gray-scale difference and temperature difference of each recognized part in the sub-model, the expansion speed and trend of cracks can be grasped in real time, providing a scientific basis for the dynamic health assessment of the bridge. Compared with traditional single static detection, dynamic monitoring can detect potential risks earlier, providing sufficient time for bridge maintenance and reinforcement. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] For the convenience of those skilled in the art to understand, the present invention will be further described below in conjunction with the accompanying drawings.

[0044] Figure 1 It is the system block diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0045] To further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following will describe in detail the specific embodiments, structures, features, and effects of the present invention in conjunction with the accompanying drawings and preferred embodiments.

[0046] Embodiment 1

[0047] Please refer to Figure 1 As shown, the present invention provides a bridge crack image detection and health monitoring system based on an unmanned aerial vehicle, including an unmanned aerial vehicle module, a computing platform, and a database;

[0048] The database performs three-dimensional modeling on the bridge to obtain a monitoring model corresponding to the bridge;

[0049] The computing platform is used to receive the monitoring model, and in the monitoring model, a plurality of position-matching points are set based on the fixed parts of the bridge, and the monitoring model is divided into multiple sub-models based on each position-matching point, and at least one set of feature points is set for each sub-model;

[0050] The position matching points are key parts of the bridge such as support points, connection points, expansion joints, etc. The feature points represent information such as the geometric shape and structural characteristics of this area. Through the geometric and texture information of the 3D model, the extracted feature points will be used as the "template" feature points of the sub-model;

[0051] The drone module plans the flight path based on each position matching point in the monitoring model. When the drone module moves along the flight path, it acquires the detection image data of the corresponding bridge and transmits the detection image data to the computing platform;

[0052] When the drone module acquires the detection image data, it adds a timestamp to the detection image data;

[0053] The computing platform intercepts the detection image data within the corresponding time period based on the matching time between the drone module and each position matching point to obtain the node image data, and the computing platform matches each node image data with the sub-model corresponding to the position matching point in the monitoring model;

[0054] The computing platform synchronizes the flight path of the drone module with the timestamp of the position matching point based on the predetermined time period of the position matching point. The computing platform first obtains the time period corresponding to each position matching point to ensure the time matching between each position matching point and the node image data;

[0055] The computing platform establishes a data set for each sub-model to collect the node image data obtained by the drone module each time, and the computing platform analyzes the time series of each data set to obtain the health data of the corresponding sub-model part of the bridge;

[0056] The method for the computing platform to match each node image data with the sub-model includes the following steps:

[0057] S1. The computer platform extracts the key points in the image from each node image data, and each key point is respectively matched with the feature points in the corresponding sub-model;

[0058] S2. The computer platform uses the feature matching algorithm to match the key points extracted from the node image data with the feature points in the corresponding sub-model;

[0059] The matching algorithm calculates the similarity according to the descriptors of the feature points and selects the best matching points;

[0060] S3. After the feature matching is completed, the computer platform will calculate the geometric transformation matrix according to the matched key points and feature points;

[0061] The computer platform corrects the image by using perspective transformation to align the feature points in the node image with the position matching points in the 3D monitoring model;

[0062] S4. According to the calculated geometric transformation matrix, the computer platform performs geometric transformation on the image to align the geometric shape of the image in the node image data with the geometric shape of the corresponding sub-model, ensuring the accurate spatial matching between the node image data and the sub-model;

[0063] The computer platform divides the node image data in the sub-model dataset into grids to generate multiple recognition parts, and uses the gray-scale difference and time-series analysis between the recognition parts to perform bridge crack detection and health monitoring. The specific steps are as follows:

[0064] Y1. The computer platform divides each node image data I into small regions of m×n according to the grid, and each small region is a recognition part R i,j , where i and j are the grid row number and column number respectively;

[0065] Y2. For each recognition part R i,j calculate its average gray-scale value G i,j , and generate the corresponding gray-scale image. The formula used is:

[0066]

[0067] In the formula, G i,j represents the average gray-scale value of the recognition part R i,j in the i-th row and j-th column, which is used to describe the overall brightness level of this grid area and reflect the characteristics of this part of the image;

[0068] w and h represent the width and height of the recognition part R i,j respectively, then w·h is the total number of pixel points in the grid area;

[0069] is the accumulation of the gray-scale values of all pixel points in the recognition part R i,j ;

[0070] R i,j (x,y) represents the gray-scale value of the pixel point at the position (x, y) in the recognition part R i,j ;

[0071] Y3. The computer platform calculates the gray-scale difference G i,j between adjacent recognition parts R i,j+1 and R i,j,j+1 in each node image data, and the formula used is: G i,j,j+1 =|G i,j -G i,j+1 |;

[0072] Y4. The computer platform performs time-series analysis on the gray-scale difference G i,j,j+1 between the same recognition parts at different time points t. The formula is:

[0073] ΔG i,j,j+1 (t) = |G i,j (t) - G i,j+1 (t)|, where t is the time point;

[0074] Y5. The computer platform evaluates the health status of the corresponding part of the bridge based on the gray - level difference and the time series of the gray - level difference between each recognition part, specifically including:

[0075] a1. The computer platform sets a difference threshold to determine whether there is an abnormal gray - level difference;

[0076] a2. Aggregate the change trend of the time series and analyze the crack propagation speed V ΔG ;

[0077] Among them, where t2 and t1 are two time points respectively;

[0078] And the computer platform generates a health score S according to the gray - level difference and the time - series change of the recognition part;

[0079] In this embodiment, where S is the health score, N is the total number of recognition parts. The lower the health score, the greater the number of cracks and the gray - level difference, and the worse the health status of the bridge;

[0080] At the same time, the calculation module transmits the node image data in each dataset to the database for backup.

[0081] Embodiment 2

[0082] A bridge crack image detection and health monitoring system based on an unmanned aerial vehicle, including an unmanned aerial vehicle module, a computing platform, and a database;

[0083] The database performs three - dimensional modeling on the bridge to obtain a monitoring model of the corresponding bridge;

[0084] The computing platform is used to receive the monitoring model, and in the monitoring model, multiple position matching points are set based on the fixed parts of the bridge. And the monitoring model is divided into multiple sub - models based on each position matching point, and at least one set of feature points is set for each sub - model;

[0085] The unmanned aerial vehicle module plans a flight path based on each position matching point in the monitoring model. And when the unmanned aerial vehicle module moves along the flight path, it acquires the detection image data of the corresponding bridge and transmits the detection image data to the computing platform;

[0086] When the unmanned aerial vehicle module acquires the detection image data, it adds a time stamp to the detection image data;

[0087] The computing platform intercepts the detected image data within the corresponding time period based on the matching time between the drone module and the matching points at each location to obtain the node image data, and the computing platform matches the sub-models between each node image data and the matching points at the corresponding positions of the monitoring model;

[0088] Based on the predetermined time period of the position matching points, the computing platform synchronizes the flight path of the drone module with the timestamps of the position matching points. The computing platform first obtains the time period corresponding to each position matching point to ensure the time matching between each position matching point and the node image data;

[0089] The computing platform establishes a data set for each sub-model to collect the node image data obtained by the drone module each time, and the computing platform analyzes the time series of each data set to obtain the health data of the corresponding sub-model part of the bridge;

[0090] The method for the computing platform to match each node image data with the sub-model includes the following steps:

[0091] S1. The computer platform extracts the key points in the image from each node image data, and each key point is respectively matched with the feature points in the corresponding sub-model;

[0092] S2. The computer platform uses a feature matching algorithm to match the key points extracted from the node image data with the feature points in the corresponding sub-model;

[0093] The matching algorithm calculates the similarity according to the descriptors of the feature points and selects the best matching points;

[0094] S3. After the feature matching is completed, the computer platform will calculate the geometric transformation matrix according to the matched key points and feature points;

[0095] The computer platform corrects the image by using perspective transformation to align the feature points in the node image with the position matching points in the three-dimensional monitoring model;

[0096] S4. According to the calculated geometric transformation matrix, the computer platform performs geometric transformation on the image to align the geometric shape of the image in the node image data with the geometric shape of the corresponding sub-model;

[0097] The computer platform divides each node image data in the sub-model data set into grids to generate multiple recognition parts, and uses the gray level difference and time series analysis between the recognition parts to perform bridge crack detection and health monitoring. The specific steps are as follows:

[0098] Y1. The computer platform divides each node image data I into small areas of m×n according to the grid, and each small area is a recognition part R i,j where i and j are the grid row number and column number respectively;

[0099] And the computer platform performs histogram equalization on the node image data of each recognition part. By redistributing the gray values of the image pixels, the gray distribution of the image becomes more uniform, enhancing the contrast. The specific steps are as follows:

[0100] Z1. The computer platform counts the number of occurrences of each gray level (0 - 255) in the node image data to generate a gray histogram. The formula used is H(g) = Count(I(x, y) = g), where g ∈ [0, 225];

[0101] In the formula, H(g) is the frequency of the gray value g, and I(x, y) is the pixel gray value at the position (x, y) in the node image data;

[0102] Z2. Accumulate the gray histogram and perform normalization processing;

[0103] The cumulative gray histogram is calculated using the cumulative probability distribution function. The formula is

[0104] In the formula, M·N is the total number of image pixels, and C(g) is the cumulative probability of pixels with gray values less than or equal to g;

[0105] The normalization process includes mapping the value of C(g) to the gray level [0, 225] to obtain the normalized gray mapping value C′(g), and C′(g) = [255·C(g)];

[0106] Z3. Replace each pixel gray value g of the node image data I(x, y) with the gray mapping value C′(g) to obtain the equalized image I′(x, y).

[0107] Y2. For each recognition part R i,j Calculate its average gray value G i,j , and generate the corresponding gray image. The formula used is:

[0108]

[0109] In the formula, G i,j represents the average gray value of the recognition part R at the i-th row and j-th column, which is used to describe the overall brightness level of the grid area and reflect the characteristics of this part of the image; i,j The average gray value of the recognition part R at the i-th row and j-th column, which is used to describe the overall brightness level of the grid area and reflect the characteristics of this part of the image;

[0110] w and h respectively represent the width and height of the recognition part R i,j , then w·h is the total number of pixel points in the grid area;

[0111] is the sum of the gray values of all pixel points within the recognition part R i,j ;

[0112] R i,j (x, y) represents the recognition part R i,j the gray value of the pixel at position (x, y) in it;

[0113] Y3. The computer platform calculates the gray difference G between adjacent recognition parts R in the node image data of each node i,j and R i,j+1 and calculates the gray difference G between adjacent recognition parts, and the formula used is: G i,j,j+1 = |G i,j,j+1 - G i,j |; i,j+1 |;

[0114] Y4. At different time points t, the computer platform performs time series analysis on the gray difference G between the same recognition parts, and the formula is: i,j,j+1 ΔG

[0115] ΔG i,j,j+1 (t) = |G i,j (t) - G i,j+1 (t)|, where t is the time point;

[0116] Y5. The computer platform evaluates the health status of the corresponding part of the bridge based on the gray difference between each recognition part and the time series of the gray difference, specifically including:

[0117] a1. The computer platform sets a difference threshold to determine whether there is an abnormal gray difference;

[0118] a2. Aggregate the change trend of the time series and analyze the crack propagation speed V ΔG ;

[0119] Among them, in the formula, t2 and t1 are two time points respectively;

[0120] And the computer platform generates a health score S according to the gray difference of the recognition part and the change of the time series;

[0121] At the same time, the calculation module transmits the node image data in each data set to the database for backup.

[0122] Example 3

[0123] A bridge crack image detection and health monitoring system based on an unmanned aerial vehicle, including an unmanned aerial vehicle module, a computing platform, and a database;

[0124] The database performs three-dimensional modeling on the bridge to obtain a monitoring model of the corresponding bridge;

[0125] The computing platform is used to receive a monitoring model. In the monitoring model, multiple position matching points are set based on the fixed parts of the bridge. And the monitoring model is divided into multiple sub-models based on each position matching point, and at least one set of feature points is set for each sub-model.

[0126] The drone module plans a flight path based on each position matching point in the monitoring model. When the drone module moves along the flight path, it acquires the detection image data of the corresponding bridge and transmits the detection image data to the computing platform.

[0127] When the drone module acquires the detection image data, it adds a timestamp to the detection image data.

[0128] And the drone module includes an infrared unit for acquiring the infrared image of the bridge. When the infrared unit acquires the infrared image of the bridge, it adds the same timestamp as the detection image data to the infrared image to ensure that the spatial positions of the infrared image and the detection image data for generating the grayscale image later are in one-to-one correspondence.

[0129] The computing platform intercepts the detection image data within the corresponding time period based on the matching time between the drone module and each position matching point to obtain the node image data, and the computing platform matches each node image data with the sub-model corresponding to the position matching point in the monitoring model.

[0130] The computing platform synchronizes the flight path of the drone module with the timestamp of the position matching point based on the predetermined time period of the position matching point. The computing platform first obtains the time period corresponding to each position matching point to ensure the time matching between each position matching point and the node image data.

[0131] The computing platform establishes a data set for each sub-model to collect the node image data obtained by the drone module each time, and the computing platform analyzes the time series of each data set to obtain the health data of the corresponding sub-model part of the bridge.

[0132] The method for the computing platform to match each node image data with the sub-model includes the following steps:

[0133] S1. The computer platform extracts the key points in the image from each node image data, and each key point is respectively matched with the feature points in the corresponding sub-model.

[0134] S2. The computer platform matches the key points extracted from the node image data with the feature points in the corresponding sub-model through a feature matching algorithm.

[0135] The matching algorithm calculates the similarity based on the descriptors of the feature points and selects the best matching points.

[0136] S3. After the feature matching is completed, the computer platform will calculate the geometric transformation matrix based on the matched key points and feature points;

[0137] S4. According to the calculated geometric transformation matrix, the computer platform will perform geometric transformation on the image to align the geometric shape of the image in the node image data with the geometric shape of the corresponding sub-model;

[0138] The computer platform will perform mesh division on each node image data in the sub-model dataset to generate multiple recognition parts, and use the gray-scale difference and time series analysis between the recognition parts to perform bridge crack detection and health monitoring. The specific steps are as follows:

[0139] Y1. The computer platform will divide each node image data I into small regions of m×n according to the grid, and each small region is a recognition part R i,j , where i and j are the grid row number and column number respectively;

[0140] Y2. For each recognition part R i,j calculate its average gray value G i,j , and generate the corresponding gray-scale image. The formula used is:

[0141]

[0142] Y3. The computer platform calculates the gray-scale difference G i,j between adjacent recognition parts R i,j+1 and R i,j,j+1 in each node image data. The formula used is: G i,j,j+1 =|G i,j -G i,j+1 |;

[0143] Y4. The computer platform also divides the infrared image into multiple infrared recognition parts according to the m×n grid, calculates the average temperature T i,j for each infrared recognition part, and calculates the temperature difference ΔT i,j,j+1 between adjacent infrared recognition parts. The formula is ΔT i,j,j+1 =|T i,j -T i,j,j+1 |;

[0144] And the computer platform combines the temperature difference between adjacent infrared recognition parts with the gray-scale difference of the corresponding node image data to form a comprehensive feature difference;

[0145] The formula for the comprehensive feature difference is ΔF i,j,j+1 =β1G i,j,j+1 +β2ΔT i,j,j+1 ;

[0146] In the formula, ΔFi,j,j+1 For the comprehensive feature differences, β1 and β2 are weight parameters, representing the contribution ratios of the gray-scale difference and the temperature difference to the detection, which are determined through experiments;

[0147] And the computer platform generates a health score S according to the time-series change of the comprehensive feature differences;

[0148] In this embodiment,

[0149] Meanwhile, the calculation module transmits the node image data in each dataset to the database for backup.

[0150] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to the above-disclosed technical content to form equivalent embodiments with equivalent changes, but as long as the technical content of the present invention is not departed from, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A bridge crack image detection and health monitoring system based on an unmanned aerial vehicle, comprising an unmanned aerial vehicle module, a computing platform and a database, characterized in that: The database transmits the monitoring model of the corresponding bridge to the computer platform; In the monitoring model of the computing platform, a plurality of position matching points are set based on the fixed parts of the bridge, and the monitoring model is divided into a plurality of sub-models based on each position matching point; At least one set of feature points is set for each sub-model; The unmanned aerial vehicle module plans a flight path based on each position matching point in the monitoring model, and when the unmanned aerial vehicle module moves along the flight path, it acquires the detection image data of the corresponding bridge and transmits the detection image data to the computing platform; The computing platform intercepts the detection image data within the corresponding time period based on the matching time between the unmanned aerial vehicle module and each position matching point to obtain the node image data, and the computing platform matches the node image data with the sub-model corresponding to the position matching point in the monitoring model; The computing platform establishes a data set for each sub-model to collect the node image data acquired by the unmanned aerial vehicle module each time, and the computing platform analyzes the time series of each data set to obtain the health data of the corresponding sub-model part of the bridge; At the same time, the computing module transmits the node image data in each data set to the database for backup.

2. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 1, characterized in that, When the unmanned aerial vehicle module acquires the detection image data, it adds a time stamp to the detection image data; The computing platform synchronizes the flight path of the unmanned aerial vehicle module with the time stamp of the position matching point based on the predetermined time period of the position matching point. The computing platform first obtains the time period corresponding to each position matching point to ensure the time matching between each position matching point and the node image data.

3. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 2, wherein, The method for the computing platform to match each node image data with the sub-model includes the following steps: S1. The computer platform extracts the key points in the image from each node image data, and each key point is respectively matched with the feature points in the corresponding sub-model; S2. The computer platform uses a feature matching algorithm to match the key points extracted from the node image data with the feature points in the corresponding sub-model; S3. After the feature matching is completed, the computer platform calculates a geometric transformation matrix according to the matched key points and feature points; The computer platform corrects the image by using perspective transformation to align the feature points in the node image with the position matching points in the three-dimensional monitoring model; S4. According to the calculated geometric transformation matrix, the computer platform performs a geometric transformation on the image to align the geometric shape of the image in the node image data with the geometric shape of the corresponding sub-model.

4. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 3, characterized in that, The computer platform divides each node image data in the sub-model data set into a grid to generate a plurality of recognition parts, and uses the gray difference and time series analysis between the recognition parts to perform bridge crack detection and health monitoring.

5. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 4, characterized in that, The steps for the computer platform to calculate the gray difference and time series analysis between each recognition part are as follows: Y1. The computer platform divides each node image data I into small regions of m×n according to the grid, and each small region is a recognition part; Y2. Calculate the average gray value for each recognized part to generate the corresponding gray image; Y3. The computer platform calculates the gray difference between adjacent recognized parts in the image data of each node; Y4. The computer platform performs time series analysis on the gray difference between the same recognized parts at different time points t; Y5. The computer platform evaluates the health status of the corresponding part of the bridge based on the gray difference between each recognized part and the time series of the gray difference; 6. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 5, wherein Before the computer platform recognizes the gray difference of each recognized part, histogram equalization is performed on the node image data of each recognized part.

7. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 5, characterized in that, The shown drone module includes an infrared unit for obtaining the infrared image of the bridge, and when the infrared unit obtains the infrared image of the bridge, a time stamp same as that of the detection image data is added to the infrared image.

8. The bridge crack image detection and health monitoring system based on an unmanned aerial vehicle according to claim 7, characterized in that, The computer platform also divides the infrared image into multiple infrared recognized parts according to the m×n grid, calculates the average temperature for each infrared recognized part, and calculates the temperature difference between adjacent infrared recognized parts; And the computer platform combines the temperature difference between adjacent infrared recognized parts with the gray difference of the corresponding node image data to form a comprehensive feature difference; The formula for the comprehensive feature difference is ΔF i,j,j+1 = β1G i,j,j+1 + β2ΔT i,j,j+1 ; where, ΔF i,j,j+1 is the comprehensive feature difference, and β1 and β2 are weight parameters; G i,j,j+1 is the gray difference of the image data of the corresponding node; ΔT i,j,j+1 is the temperature difference between adjacent infrared recognition parts; And the computer platform generates a health score according to the time series change of the comprehensive feature difference.

Images