Bridge health cloud monitoring system based on computer vision

Abstract

The computer vision-based bridge health cloud monitoring system includes a visual terminal and a cloud server; the visual terminal sends the pre-processed image data to the cloud server via WAN and 5G; the cloud server performs three-dimensional reconstruction based on the image data to obtain a three-dimensional reconstruction model, and compares it with the three-dimensional design drawing of the bridge to obtain a three-dimensional comparison result; according to the three-dimensional comparison result, the current health status of the bridge is diagnosed. The present invention uses computer vision technology to collect image data of the bridge, and its installation and maintenance are much more convenient than traditional acquisition terminals based on mechanical electronic sensors, acoustic vibration sensors and photoelectric sensors; each visual terminal can cover an area, while the sensor can only be placed at a fixed point, so the installation workload is much less than the latter; the deformation of the bridge can be displayed in the form of a three-dimensional image, which is more convenient and accurate for bridge health diagnosis compared with the data collected by the sensor.

Description

Technical Field

[0001] The present invention relates to the field of computer vision, and in particular to a bridge health cloud monitoring system based on computer vision. Background Art

[0002] With the continuous development of computer vision technology, especially the development of machine learning and deep learning in recent years, computer vision for bridge monitoring has become possible. The country's emphasis on infrastructure construction and maintenance and the development of technology require a more advanced technology. The bridge health cloud monitoring system based on computer vision can provide users with a new cloud monitoring platform through computer vision technology, supplemented by other traditional monitoring methods and expert systems, using image analysis, deep learning and 3D reconstruction technology and 3D design data of bridges.

[0003] Existing technologies are mainly based on traditional mechanical electronic sensors, acoustic vibration sensors and photoelectric sensors, which are connected to the bridge body in a contact manner. They are troublesome and risky to install and difficult to maintain. Some technologies store data locally, and some technologies transmit data to cloud servers, but they cannot collect images and three-dimensional data of the bridge, and cannot provide a visual, re-analyzable monitoring and early warning platform for the background. Currently, many bridges also have cameras, but they only use surveillance cameras for security monitoring, not image analysis. Summary of the invention

[0004] In view of the above problems existing in the prior art, the present invention provides a bridge health cloud monitoring system based on computer vision, which can detect and warn bridges through computer vision technology.

[0005] The technical solution of the present invention is as follows:

[0006] The computer vision-based bridge health cloud monitoring system includes a visual terminal and a cloud server; the visual terminal is connected to the cloud server via WAN and 5G;

[0007] The visual terminal includes an industrial camera and lens, a microprocessor, and a laser module;

[0008] The laser module is used for on-site illumination and improving imaging quality;

[0009] The software of the microprocessor includes an acquisition module, a preprocessing module and a transmission module;

[0010] The acquisition module acquires image data of the measured parts of the bridge through an industrial camera and a lens; the measured parts include but are not limited to: cables, bridge arches, and bridge bodies;

[0011] The preprocessing module preprocesses the image data from the acquisition module; the preprocessing includes but is not limited to: image enhancement, image denoising and image alignment;

[0012] The transmission module sends the image data from the pre-processing module to the cloud server;

[0013] The cloud server includes background software and a database; the background software includes a receiving module, a storage module, an analysis module, a diagnosis module, an early warning module, a management module, a query module, and a service module;

[0014] The receiving module receives image data of multiple visual terminals in parallel;

[0015] The functions of the storage module include: saving the image data from the visual terminal into a database; saving the three-dimensional reconstruction model of the measured part into a database; saving the three-dimensional comparison result between the three-dimensional reconstruction model of the measured part and the three-dimensional design drawing of the bridge into a database;

[0016] The analysis module uses an artificial intelligence neural network and concurrent computing mode to calculate the corresponding characteristic values ​​according to the image data, and performs three-dimensional reconstruction of the measured part to obtain a three-dimensional reconstruction model and compare it with the three-dimensional design drawing of the bridge to obtain a three-dimensional comparison result;

[0017] The diagnosis module classifies and regresses the 3D reconstruction model and the 3D comparison results based on machine learning technology, and performs real-time diagnosis of the health status of the bridge;

[0018] The early warning module uses machine learning technology to combine the historical data of the 3D reconstruction model and 3D comparison results with the professional knowledge of the expert group to achieve early health warning of the bridge;

[0019] The functions of the management module include but are not limited to: setting user permissions at all levels; importing the three-dimensional design drawings of the bridge; marking the measured parts of the bridge;

[0020] The query module provides users with a web-based data query system;

[0021] The functions of the service module include but are not limited to: providing offline consultation and real-time online consultation, connecting with expert teams, solving difficult problems in bridge maintenance, and providing final repair and maintenance solutions.

[0022] Furthermore, the algorithm used for image enhancement is CLAHE, i.e., contrast-limited adaptive histogram equalization algorithm, and the specific steps are as follows:

[0023] S21: Divide the image into several sub-blocks, count the pixels of each sub-block, and obtain a histogram;

[0024] S22: Set an upper limit ClipLimit for the histogram, and sum the part of the histogram that is higher than the upper limit ClipLimit, and the result is totalExcess;

[0025] S23: Divide totalExcess equally to all gray levels in the histogram, and calculate the height L = totalExcess / N of the overall increase in the histogram caused by this operation, where N is the number of gray levels;

[0026] S24: Calculate the difference between the upper limit ClipLimit and the height L: Upper = ClipLimit-L;

[0027] S25: With ClipLimit and Upper as the boundaries, the histogram is processed as follows: if the amplitude is higher than ClipLimit, it is directly set to ClipLimit; if the amplitude is between Upper and ClipLimit, it is filled to ClipLimit; if the amplitude is lower than Upper, L pixels are directly filled.

[0028] Furthermore, the algorithm used for image denoising is a non-local median algorithm, and the specific steps are as follows:

[0029] The image is divided into several sub-blocks, one of the sub-blocks is randomly selected, the pixels of this sub-block are compared with the pixels of other areas in the entire image, and the similarity between the two is calculated. The higher the similarity, the greater the weight obtained; if the similarity is lower than a certain threshold, the weight is zero; these similar pixel values ​​are weighted and summed according to the normalized weights to obtain the denoised image.

[0030] Furthermore, the algorithm used for image alignment is a global image alignment method, and the specific steps are as follows:

[0031] S41: extracting features from the image, and extracting image feature edges using an edge operator;

[0032] S42: rotating the original feature image to form a series of feature images; generating a number of positive and negative rotation feature images according to the pre-configured rotation angle and precision configuration;

[0033] S43: Reduce the dimension of the feature sub-array, take the length and width of 1 / 4 of the original image, and obtain 1 / 16 of the original image, and perform downsampling using the mean method;

[0034] S44: Calculate the normalized cross-correlation coefficient for all features of the feature subarray to obtain a coarse-grained matching image;

[0035] S45: Finding adjacent feature images in the high-precision space to perform high-precision feature image matching;

[0036] S46: Rotate, translate and shear the image according to the image matching result to obtain an aligned image.

[0037] Furthermore, the edge operators include but are not limited to Canny and Sobel.

[0038] Furthermore, the number of the forward and reverse rotation feature images is 20.

[0039] Furthermore, the value of the rotation angle is 2°, including both positive and negative directions.

[0040] Furthermore, the value of the precision configuration is 0.2°.

[0041] Further, the use process includes the following steps:

[0042] S1: Select the measurement location of the bridge;

[0043] S2: Install the visual terminal at a suitable location to ensure that the visual terminal can collect image data of the measurement site, and the same measurement site is covered by the collection areas of at least two visual terminals;

[0044] S3: The visual terminal collects image data of the measurement part in real time or at a fixed time, and pre-processes the image data;

[0045] S4: The visual terminal sends the pre-processed image data to the cloud server via WAN and 5G;

[0046] S5: The receiving module of the cloud server receives the image data from multiple visual terminals; the storage module of the cloud server stores the image data in a database in a segmented and encrypted manner;

[0047] S6: The analysis module of the cloud server reads the image data in the database and calculates the corresponding characteristic values ​​according to the processing mode, algorithm and monitoring area set by the customer; the characteristic values ​​include but are not limited to: crack size and deformation, bridge main vibration frequency, cable vibration frequency, main tower beam deformation;

[0048] S7: The analysis module of the cloud server performs three-dimensional reconstruction based on the image data to obtain a three-dimensional reconstruction model; the three-dimensional reconstruction model is compared with the three-dimensional design drawing of the bridge to obtain a three-dimensional comparison result;

[0049] S8. The diagnosis module of the cloud server diagnoses the current health status of the bridge based on the three-dimensional comparison results obtained by the analysis module using big data analysis and machine learning technology;

[0050] S9. The early warning module of the cloud server uses machine learning technology to issue early warnings on bridge health based on current and historical diagnostic results, providing a reference for bridge maintenance.

[0051] Furthermore, the specific steps of step S7 are as follows:

[0052] S61: Determine a three-dimensional space coordinate system of the bridge, i.e., a bridge coordinate system, according to the three-dimensional design drawing of the bridge;

[0053] S62: determining the coordinates of the visual terminal in the bridge coordinate system according to the three-dimensional design drawing of the bridge and the actual installation position of the visual terminal;

[0054] S63: Reading the image data of each visual terminal at the same time from the database, constructing the three-dimensional contour of the measurement part according to the binocular vision stereo imaging principle, obtaining the three-dimensional reconstruction model and saving it in the database;

[0055] S64: Mapping the three-dimensional reconstructed model of the measurement part into the bridge coordinate system through the coordinate space conversion relationship with the visual terminal;

[0056] S65: matching the three-dimensional reconstructed model of the measured part with the three-dimensional design drawing of the bridge, calculating the deviation between the two, and saving them in a database; matching parameters include but are not limited to: position, length, and angle of key points;

[0057] S66: Search the database, obtain the deviation of the measurement part within a period of time, and calculate the variation trend of the deviation; when the size of the deviation or the variation trend exceeds the warning threshold, the warning module will remind or alarm.

[0058] The beneficial technical effects of the present invention are:

[0059] (1) The image data of bridges is collected by computer vision technology, which is much easier to install and maintain than traditional data collection terminals based on mechanical electronic sensors, acoustic vibration sensors, and photoelectric sensors;

[0060] (2) Each visual terminal can cover an area, while sensors can only be placed at fixed points, so the installation workload is much less than the latter;

[0061] (3) The deformation of the bridge can be displayed in the form of a three-dimensional image, which is more convenient and accurate for bridge health diagnosis compared with the data collected by sensors. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the structure of the visual terminal;

[0063] Figure 2 It is a schematic diagram of the structure of the cloud server;

[0064] Figure 3 It is a schematic diagram of the algorithm for image denoising;

[0065] Figure 4 It is a schematic diagram of the algorithm for image alignment;

[0066] Figure 5 This is the installation diagram of the visual terminal. Figure 1 ;

[0067] Figure 6 This is the installation diagram of the visual terminal. Figure 2 . DETAILED DESCRIPTION

[0068] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The monitoring object of the embodiment is a cable-stayed bridge with a length of 930 meters, a bridge deck tower height of 28 meters, and a total of 100 cables.

[0070] The solution of the embodiment is: a total of 20 visual terminals and one cloud server; 10 visual terminals collect cable data, 2 visual terminals collect tower beam data, and 8 visual terminals are placed in the beam box to collect data on key connection points and cracks.

[0071] The visual terminal is connected to the cloud server via WAN and 5G.

[0072] like Figure 1 As shown, the visual terminal includes an industrial camera and lens, a microprocessor, and a laser module. The laser module is used for on-site lighting and improving imaging quality.

[0073] The software of the microprocessor includes acquisition module, preprocessing module and transmission module, and its functions are as follows:

[0074] (1) The acquisition module collects image data of the measured part of the bridge through an industrial camera and lens;

[0075] (2) The preprocessing module preprocesses the image data from the acquisition module, mainly including: image enhancement, image denoising and image alignment.

[0076] The algorithm used for image enhancement is CLAHE, which is the contrast-limited adaptive histogram equalization algorithm. The specific steps are as follows:

[0077] S21: Divide the image into several sub-blocks, count the pixels of each sub-block, and obtain a histogram;

[0078] S22: Set an upper limit ClipLimit for the histogram, and sum the part of the histogram that is higher than the upper limit ClipLimit, and the result is totalExcess;

[0079] S23: Divide totalExcess equally to all gray levels in the histogram, and calculate the height L = totalExcess / N of the overall increase in the histogram caused by this operation, where N is the number of gray levels;

[0080] S24: Calculate the difference between the upper limit ClipLimit and the height L: Upper = ClipLimit-L;

[0081] S25: With ClipLimit and Upper as the boundaries, the histogram is processed as follows: if the amplitude is higher than ClipLimit, it is directly set to ClipLimit; if the amplitude is between Upper and ClipLimit, it is filled to ClipLimit; if the amplitude is lower than Upper, L pixels are directly filled.

[0082] The algorithm used for image denoising is the non-local median algorithm, such as Figure 3 As shown, the specific steps are as follows:

[0083] The image is divided into several sub-blocks, one of the sub-blocks is randomly selected, the pixels of this sub-block are compared with the pixels of other areas in the entire image, and the similarity between the two is calculated. The higher the similarity, the greater the weight obtained; if the similarity is lower than a certain threshold, the weight is zero; these similar pixel values ​​are weighted and summed according to the normalized weights to obtain the denoised image.

[0084] The algorithm used for image alignment is the global image alignment method, such as Figure 4 As shown, the specific steps are as follows:

[0085] S41: extract features from the image and use edge operators to extract image feature edges; edge operators include Canny and Sobel;

[0086] S42: rotating the original feature image to form a series of feature images, including the rotated feature image; generating 20 positive and negative rotated feature images according to the pre-configured rotation angle and precision configuration; the value of the rotation angle is 2°, including both positive and negative directions; the value of the precision configuration is 0.2°;

[0087] S43: Reduce the dimension of the feature sub-array, and take the length and width of 1 / 4 of the original image, that is, 1 / 16 of the original image, and down-sample by the mean method;

[0088] S44: Calculate the normalized cross-correlation coefficient for all features of the feature subarray to obtain a coarse-grained matching image;

[0089] S45: Finding adjacent feature images in the high-precision space to perform high-precision feature image matching;

[0090] S46: Rotate, translate and shear the image according to the image matching result to obtain an aligned image.

[0091] (3) The transmission module sends the image data from the preprocessing module to the cloud server.

[0092] like Figure 2 As shown, the cloud server includes background software and database; the background software includes receiving module, storage module, analysis module, diagnosis module, warning module, management module, query module, and service module. The functions of each module are as follows:

[0093] (1) The receiving module receives image data from all visual terminals in parallel;

[0094] (2) The functions of the storage module include: saving the image data from the visual terminal into a database; saving the three-dimensional reconstruction model of the measured part into a database; saving the three-dimensional comparison result between the three-dimensional reconstruction model of the measured part and the three-dimensional design drawing of the bridge into a database;

[0095] (3) The analysis module uses an artificial intelligence neural network and concurrent computing mode to calculate the corresponding characteristic values ​​based on the image data, and performs three-dimensional reconstruction of the measured part to obtain a three-dimensional reconstruction model and compare it with the three-dimensional design drawing of the bridge to obtain a three-dimensional comparison result;

[0096] (4) The diagnosis module uses machine learning technology to classify and regress the 3D reconstruction model and 3D comparison results to perform real-time diagnosis of the health status of the bridge;

[0097] (5) The early warning module uses machine learning technology to combine historical data with the professional knowledge of the expert group to achieve early health warning of the bridge;

[0098] (6) The functions of the management module include but are not limited to: setting user permissions at all levels, importing three-dimensional design drawings of bridges, and marking the measurement locations of bridges;

[0099] (7) The query module provides users with a web-based data query system;

[0100] (8) The functions of the service module include: providing offline consultation and real-time online consultation, connecting with expert teams, solving difficult problems in bridge maintenance, and providing final repair and maintenance solutions.

[0101] The use process of this embodiment includes the following steps:

[0102] S1: Select the measurement parts of the bridge, such as tower beams, cables, bridge section connections and some cracks with potential risks, etc.

[0103] S2: Figure 5 , Figure 6 As shown, the visual terminal is installed at a suitable position to ensure that the visual terminal can collect image data of the measurement part, and the same measurement part is covered by the collection areas of at least two visual terminals;

[0104] S3: The visual terminal collects image data of the measurement part in real time or at a fixed time, and pre-processes the image data;

[0105] S4: The visual terminal sends the pre-processed image data to the cloud server via WAN and 5G;

[0106] S5: The receiving module of the cloud server receives the image data from multiple visual terminals; the storage module of the cloud server stores the image data in a database in a segmented and encrypted manner;

[0107] S6: The analysis module of the cloud server reads the image data in the database and calculates the corresponding characteristic value according to the processing mode, algorithm and monitoring area set by the customer;

[0108] S7: The analysis module of the cloud server performs 3D reconstruction based on the image data, obtains a 3D reconstruction model and compares it with the 3D design drawing of the bridge to obtain a 3D comparison result. The specific steps are as follows:

[0109] (1) According to the three-dimensional design drawing of the bridge, determine the three-dimensional spatial coordinate system of the bridge, that is, the bridge coordinate system;

[0110] (2) Determine the coordinates of the visual terminal in the bridge coordinate system based on the three-dimensional design drawing of the bridge and the actual installation position of the visual terminal;

[0111] (3) Extracting the image data of each visual terminal at the same time, constructing the three-dimensional contour of the measurement part according to the principle of binocular vision stereo imaging, obtaining the three-dimensional reconstruction model and saving it in the database;

[0112] (4) Mapping the three-dimensional reconstructed model of the measured part into the bridge coordinate system through the coordinate space conversion relationship with the visual terminal;

[0113] (5) Match the 3D reconstructed model of the measured part with the 3D design drawing of the bridge (including matching of key point positions, lengths, angles and other parameters), calculate the deviation between the two, and save it in the database;

[0114] (6) Search the database to obtain the deviation of the measurement part within a period of time and calculate the change trend of the deviation; when the size of the deviation or the change trend exceeds the warning threshold, the warning module will remind or alarm.

[0115] S8. The diagnosis module of the cloud server diagnoses the current health status of the bridge based on the three-dimensional comparison results obtained by the analysis module, using big data analysis and machine learning technology.

[0116] S9. The early warning module of the cloud server uses machine learning technology to issue early warnings on bridge health based on current and historical diagnostic results, providing a reference for bridge maintenance.

[0117] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and implementation modes, and can be fully applicable to various fields suitable for the present invention. For those familiar with the art, for those of ordinary skill in the art, various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to specific details.

Claims

1. The bridge health cloud monitoring system based on computer vision is characterized by: It includes a visual terminal and a cloud server; the visual terminal is connected to the cloud server via WAN and 5G; The visual terminal includes an industrial camera and lens, a microprocessor, and a laser module; The laser module is used for on-site illumination and improving imaging quality; The software of the microprocessor includes an acquisition module, a preprocessing module and a transmission module; The acquisition module acquires image data of the measured part of the bridge through an industrial camera and a lens; The measuring parts include: cables, bridge arches, and bridge bodies; The preprocessing module preprocesses the image data from the acquisition module; the preprocessing includes: image enhancement, image denoising and image alignment; The transmission module sends the image data from the pre-processing module to the cloud server; The cloud server includes background software and a database; the background software includes a receiving module, a storage module, an analysis module, a diagnosis module, an early warning module, a management module, a query module, and a service module; The receiving module receives image data of multiple visual terminals in parallel; The functions of the storage module include: saving the image data from the visual terminal into a database; saving the three-dimensional reconstruction model of the measured part into a database; saving the three-dimensional comparison result between the three-dimensional reconstruction model of the measured part and the three-dimensional design drawing of the bridge into a database; The analysis module uses an artificial intelligence neural network and concurrent computing mode to calculate the corresponding characteristic values ​​according to the image data, and performs three-dimensional reconstruction of the measured part to obtain a three-dimensional reconstruction model and compare it with the three-dimensional design drawing of the bridge to obtain a three-dimensional comparison result; The diagnosis module classifies and regresses the 3D reconstruction model and the 3D comparison results based on machine learning technology, and performs real-time diagnosis of the health status of the bridge; The early warning module uses machine learning technology to combine the historical data of the 3D reconstruction model and 3D comparison results with the professional knowledge of the expert group to achieve early health warning of the bridge; The functions of the management module include: setting user permissions at all levels; importing the three-dimensional design drawings of the bridge; marking the measured parts of the bridge; The query module provides users with a web-based data query system; The functions of the service module include: providing offline consultation and real-time online consultation, connecting with expert teams, solving difficult problems in bridge maintenance, and providing final repair and maintenance solutions; The usage process includes the following steps: S1: Select the measurement location of the bridge; S2: Install the visual terminal at a suitable location to ensure that the visual terminal can collect image data of the measurement site, and the same measurement site is covered by the collection areas of at least two visual terminals; S3: The visual terminal collects image data of the measurement part in real time or at a fixed time, and pre-processes the image data; S4: The visual terminal sends the pre-processed image data to the cloud server via WAN and 5G; S5: The receiving module of the cloud server receives the image data from multiple visual terminals; the storage module of the cloud server stores the image data in a database in a segmented and encrypted manner; S6: The analysis module of the cloud server reads the image data in the database and calculates the corresponding characteristic values ​​according to the processing mode, algorithm and monitoring area set by the customer; the characteristic values ​​include: crack size and deformation, bridge main vibration frequency, cable vibration frequency, main tower beam deformation; S7: The analysis module of the cloud server performs three-dimensional reconstruction based on the image data to obtain a three-dimensional reconstruction model; the three-dimensional reconstruction model is compared with the three-dimensional design drawing of the bridge to obtain a three-dimensional comparison result; S8. The diagnosis module of the cloud server diagnoses the current health status of the bridge based on the three-dimensional comparison results obtained by the analysis module using big data analysis and machine learning technology; S9. The early warning module of the cloud server uses machine learning technology to warn of bridge health based on current and historical diagnosis results, providing a reference for bridge maintenance; The specific steps of step S7 are as follows: S61: Determine a three-dimensional space coordinate system of the bridge, i.e., a bridge coordinate system, according to the three-dimensional design drawing of the bridge; S62: determining the coordinates of the visual terminal in the bridge coordinate system according to the three-dimensional design drawing of the bridge and the actual installation position of the visual terminal; S63: Reading the image data of each visual terminal at the same time from the database, constructing the three-dimensional contour of the measurement part according to the binocular vision stereo imaging principle, obtaining the three-dimensional reconstruction model and saving it in the database; S64: Mapping the three-dimensional reconstructed model of the measurement part into the bridge coordinate system through the coordinate space conversion relationship with the visual terminal; S65: matching the three-dimensional reconstruction model of the measured part with the three-dimensional design drawing of the bridge, calculating the deviation between the two, and saving them in a database; the matching parameters include: the position, length, and angle of the key points; S66: Search the database, obtain the deviation of the measurement part within a period of time, and calculate the variation trend of the deviation; when the size of the deviation or the variation trend exceeds the warning threshold, the warning module will remind or alarm.

2. The computer vision-based bridge health cloud monitoring system according to claim 1, characterized in that: The algorithm used for image enhancement is CLAHE, i.e., contrast-limited adaptive histogram equalization algorithm. The specific steps are as follows: S21: Divide the image into several sub-blocks, count the pixels of each sub-block, and obtain a histogram; S22: Set an upper limit ClipLimit for the histogram, and sum the part of the histogram that is higher than the upper limit ClipLimit, and the result is totalExcess; S23: Divide totalExcess evenly to all gray levels in the histogram, and find the height L=totalExcess / N that causes the overall rise of the histogram, where N is the number of gray levels; S24: Calculate the difference between the upper limit ClipLimit and the height L: Upper = ClipLimit-L; S25: With ClipLimit and Upper as the boundaries, the histogram is processed as follows: if the amplitude is higher than ClipLimit, it is directly set to ClipLimit; if the amplitude is between Upper and ClipLimit, it is filled to ClipLimit; if the amplitude is lower than Upper, L pixels are directly filled.

3. The computer vision-based bridge health cloud monitoring system according to claim 1, characterized in that: The algorithm used for image denoising is the non-local median algorithm, and the specific steps are as follows: The image is divided into several sub-blocks, one of the sub-blocks is randomly selected, the pixels of this sub-block are compared with the pixels of other areas in the entire image, and the similarity between the two is calculated. The higher the similarity, the greater the weight obtained; if the similarity is lower than a certain threshold, the weight is zero; these similar pixel values ​​are weighted and summed according to the normalized weights to obtain the denoised image.

4. The computer vision-based bridge health cloud monitoring system according to claim 1, characterized in that: The algorithm used for image alignment is a global image alignment method, and the specific steps are as follows: S41: extracting features from the image, and extracting image feature edges using an edge operator; S42: rotating the original feature image to form a series of feature images; generating a number of positive and negative rotation feature images according to the pre-configured rotation angle and precision configuration; S43: Reduce the dimension of the feature sub-array, take the length and width of 1 / 4 of the original image, and obtain 1 / 16 of the original image, and perform downsampling using the mean method; S44: Calculate the normalized cross-correlation coefficient for all features of the feature subarray to obtain a coarse-grained matching image; S45: Finding adjacent feature images in the high-precision space to perform high-precision feature image matching; S46: Rotate, translate and shear the image according to the image matching result to obtain an aligned image.

5. The computer vision-based bridge health cloud monitoring system according to claim 4 is characterized in that: The edge operators include Canny and Sobel.

6. The computer vision-based bridge health cloud monitoring system according to claim 4, characterized in that: The number of the forward and reverse rotation characteristic images is 20.

7. The computer vision-based bridge health cloud monitoring system according to claim 4 is characterized in that: The value of the rotation angle is 2°, including both positive and negative directions.

8. The computer vision-based bridge health cloud monitoring system according to claim 4, characterized in that: The value of the accuracy configuration is 0.2°.

Images