Crack detection method for concrete dam

Abstract

The invention provides a crack detection method for a concrete dam, and relates to the technical field of image processing. The method comprises the following steps: acquiring a historical image of a concrete dam and a real-time image of a to-be-detected concrete dam; performing denoising processing on the historical image and the real-time image of the concrete dam, performing crack identification after obtaining the denoised historical image and the denoised real-time image, performing crack feature analysis after obtaining a historical crack region image and a real-time crack region image, and obtaining historical crack feature data and real-time crack feature data; constructing a crack recognition model according to the marked size information of the historical crack feature data; and inputting the real-time crack characteristic data into the crack identification model to obtain real-time crack size information and then carrying out crack detection so as to obtain a real-time crack detection result. Therefore, by means of the method, the detection period of the cracks of the concrete dam can be shortened, workers do not need to check the cracks one by one, and potential safety hazards and the working intensity of the workers are reduced.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method for detecting cracks in a concrete dam. Background Art

[0002] With the gradual development of informatization and automation, machine vision technology and drone technology have been widely used in various engineering projects. Among them, water conservancy hub facilities projects require regular maintenance and operation, and crack detection is one of the essential projects. Based on the morphology and location characteristics of cracks, the inherent damage and potential causes of cracks can be inferred, which can provide reasonable guidance for structural risk assessment.

[0003] Existing crack detection of concrete dams usually adopts slow manual inspection. However, when the concrete dam is too long, the manual inspection cycle is too long, which increases the workload of the staff. In addition, concrete dams are often built in dangerous locations, and the staff may be injured during the inspection. Summary of the invention

[0004] The purpose of the embodiment of the present application is to provide a method for detecting cracks in a concrete dam, so as to solve the problem that the existing concrete dam crack detection proposed in the above background technology adopts slow manual inspection. When the concrete dam is too long, the manual inspection cycle is too long, which increases the workload of the staff. In addition, concrete dams are often built in dangerous locations, and the staff may be injured during the inspection.

[0005] To achieve the above purpose, the technical solution adopted in this application is:

[0006] The first aspect of the present application provides a method for detecting cracks in a concrete dam, comprising:

[0007] Obtain historical images of concrete dams and real-time images of concrete dams to be inspected;

[0008] De-noising the concrete dam historical image and the real-time image to obtain a de-noised historical image and a de-noised real-time image;

[0009] Perform crack identification on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image;

[0010] Performing crack feature analysis on the historical crack region image and the real-time crack region image respectively to obtain historical crack feature data corresponding to the historical crack region image and real-time crack feature data corresponding to the real-time crack region image;

[0011] Constructing a crack recognition model according to the annotated dimension information corresponding to the historical crack characteristic data;

[0012] Inputting the real-time crack characteristic data into the crack identification model to obtain real-time crack size information;

[0013] The concrete dam to be detected is subjected to crack detection according to the real-time crack size information to obtain a real-time crack detection result of the concrete dam to be detected, wherein the real-time crack detection result is used to indicate whether the crack of the concrete dam to be detected is a dangerous crack or a normal crack.

[0014] Optionally, the denoising the concrete dam historical image and the real-time image to obtain the denoised historical image and the denoised real-time image includes:

[0015] Decomposing the high-frequency information and the low-frequency information of the historical image of the concrete dam to obtain the historical high-frequency information and the historical low-frequency information corresponding to the historical image of the concrete dam;

[0016] Respectively reconstructing the historical high-frequency information and the historical low-frequency information by high-frequency pixels and low-frequency pixels to obtain a reconstructed historical high-frequency component and a reconstructed historical low-frequency component;

[0017] Aggregating the reconstructed historical high-frequency component and the reconstructed historical low-frequency component to generate the denoised historical image;

[0018] Decomposing the high-frequency information and the low-frequency information of the real-time image to obtain the real-time high-frequency information and the real-time low-frequency information corresponding to the real-time image;

[0019] Respectively reconstructing the high-frequency pixels and the low-frequency pixels of the real-time high-frequency information and the real-time low-frequency information to obtain a reconstructed real-time high-frequency component and a reconstructed real-time low-frequency component;

[0020] The reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component are aggregated to generate the denoised real-time image.

[0021] Optionally, the performing crack identification on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image includes:

[0022] Respectively performing crack rectangular frame recognition on the denoised historical image and the denoised real-time image to obtain a historical crack rectangular frame and a real-time crack rectangular frame;

[0023] Performing tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image, respectively, to obtain an angle-corrected historical image and an angle-corrected real-time image;

[0024] Screening the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image to obtain a target historical rectangular frame and a target real-time rectangular frame;

[0025] According to the target historical rectangular frame and the target real-time rectangular frame, the historical crack region image and the real-time crack region image are determined from the denoised historical image and the denoised real-time image respectively.

[0026] Optionally, the performing crack rectangular frame recognition on the denoised historical image and the denoised real-time image respectively to obtain a historical crack rectangular frame and a real-time crack rectangular frame comprises:

[0027] Performing edge detection on the denoised historical image and the denoised real-time image respectively to obtain a historical crack edge and a real-time crack edge;

[0028] According to the historical crack edge and the real-time crack edge, the denoised historical image and the denoised real-time image are processed respectively using a first morphological closing algorithm to obtain a historical passage area image in the denoised historical image and a real-time passage area image in the denoised real-time image;

[0029] Using the second morphological closing algorithm, single line interference removal is performed on the historical path area image and the real-time path area image;

[0030] Filling holes in the historical path region image and the real-time path region image after interference removal respectively;

[0031] The minimum enclosing rectangles are calculated according to each white area in the historical path region image after the hole supplementation and each white area in the real-time path region image, respectively, as the historical crack rectangular frame and the real-time crack rectangular frame.

[0032] Optionally, the performing tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image respectively to obtain the angle-corrected historical image and the angle-corrected real-time image comprises:

[0033] According to the inclination angle of the historical crack rectangular frame and the inclination angle of the real-time crack rectangular frame, respectively counting the first angle distribution of the historical crack rectangular frame within a preset right angle range and the second angle distribution of the real-time crack rectangular frame within the preset right angle range;

[0034] Determine, from the first angle distribution and the second angle distribution, respectively, a first target tilt angle and a second target tilt angle whose angle distribution satisfies a preset distribution concentration condition;

[0035] According to the first target tilt angle and the second target tilt angle, the tilt angles of the historical crack rectangular frame and the real-time crack rectangular frame are respectively corrected to obtain the angle-corrected historical image and the angle-corrected real-time image.

[0036] Optionally, the filtering of the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image to obtain a target historical rectangular frame and a target real-time rectangular frame includes:

[0037] Determine, from all the historical crack rectangular frames in the angle-corrected historical image, the historical candidate crack rectangular frames that are in the zero-degree direction or the ninety-degree direction and whose errors are within a preset angle range, and screen the historical candidate crack rectangular frames to obtain the target historical rectangular frame;

[0038] From all the real-time crack rectangular frames in the historical image after angle correction, determine the real-time candidate crack rectangular frames that are in the zero degree direction or the ninety degree direction and whose errors are within the preset angle range, and screen the real-time candidate crack rectangular frames to obtain the target real-time rectangular frame.

[0039] Optionally, the historical crack characteristic data includes: historical crack width characteristic data, historical crack length characteristic data, and historical crack depth data;

[0040] The step of constructing a crack recognition model according to the annotated dimension information corresponding to the historical crack characteristic data includes:

[0041] The crack recognition model is constructed according to the annotated width dimension information corresponding to the historical crack width characteristic data, the annotated length dimension information corresponding to the historical crack length characteristic data, and the annotated depth dimension information corresponding to the historical crack depth data.

[0042] Optionally, the real-time crack characteristic data includes: real-time crack width characteristic data, real-time crack length characteristic data, and real-time crack depth data; the real-time crack size information includes: real-time crack width size information, real-time crack length size information, and real-time crack depth size information;

[0043] The step of inputting the real-time crack characteristic data into the crack identification model to obtain real-time crack size information includes:

[0044] Inputting the real-time crack width characteristic data into the crack recognition model to obtain the real-time crack width dimension information;

[0045] Inputting the real-time crack length characteristic data into the crack recognition model to obtain the real-time crack length size information;

[0046] The real-time crack depth characteristic data is input into the crack recognition model to obtain the real-time crack depth dimension information.

[0047] Optionally, performing crack detection on the concrete dam to be detected according to the real-time crack size information to obtain a real-time crack detection result of the concrete dam to be detected includes:

[0048] Calculating real-time crack space size information according to the real-time crack width dimension information, the real-time crack length dimension information and the real-time crack depth dimension information;

[0049] If the real-time crack space size information is greater than or equal to a preset crack space threshold, a first real-time crack detection result is generated to indicate that the crack of the concrete dam to be detected is a dangerous crack;

[0050] If the real-time crack space size information is smaller than the preset crack space threshold, a second real-time crack detection result is generated to indicate that the crack of the concrete dam to be detected is a normal crack.

[0051] Optionally, the method further comprises:

[0052] If the first target crack in the concrete dam to be detected is a normal crack, crack detection is performed on the concrete dam to be detected for a continuous preset time period to obtain first crack space size information of the first target crack after the preset time period;

[0053] Calculating the crack expansion rate of the first target crack in the preset time period according to the second crack space size information of the first target crack before the preset time period;

[0054] Calculating an average crack expansion rate of the concrete dam to be inspected according to the crack expansion rates of all the first target cracks;

[0055] If the average crack expansion rate is greater than or equal to a preset crack expansion rate threshold, it is determined that the first target crack of the concrete dam to be inspected is in an abnormal expansion state;

[0056] If the average crack expansion rate is less than the preset crack expansion rate threshold, it is determined that all the first target cracks in the concrete dam to be inspected are in a normal expansion state.

[0057] The beneficial effects of the embodiments of the present application include, for example:

[0058] The present application provides a method for detecting cracks in a concrete dam, which obtains a historical image of a concrete dam and a real-time image of a concrete dam to be detected; then denoises the historical image and the real-time image of the concrete dam to obtain a denoised historical image and a denoised real-time image; and performs crack identification on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image, so as to perform denoising analysis and correction processing on the historical image of the concrete dam and the real-time image of the concrete dam to be detected, so that the historical image of the concrete dam and the real-time image data of the concrete dam to be detected are in the same standard, which is referenceable and researchable; the historical crack region image and the real-time crack region image are analyzed and corrected. The crack feature analysis is performed separately to obtain the historical crack feature data corresponding to the historical crack area image and the real-time crack feature data corresponding to the real-time crack area image; the crack recognition model is constructed according to the annotated size information corresponding to the historical crack feature data, that is, a crack recognition model that meets the requirements is constructed through a large amount of historical crack feature data; the real-time crack feature data is then input into the crack recognition model to obtain the real-time crack size information, so that the staff can grasp the real-time crack size information in real time; and the crack detection is performed on the concrete dam to be detected according to the real-time crack size information, and the real-time crack detection result of the concrete dam to be detected is obtained, which is used to indicate whether the crack of the concrete dam to be detected is a dangerous crack or a normal crack. Therefore, the present application can realize the automatic detection of cracks in concrete dams, shorten the detection cycle of cracks in concrete dams, and eliminate the need for staff to check cracks one by one, thereby reducing safety hazards and the work intensity of staff, and at the same time, it can also meet the needs of staff. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0060] Figure 1 A structural schematic diagram of a crack detection system for a concrete dam provided in an embodiment of the present application;

[0061] Figure 2 A schematic diagram of the structure of a computer device provided in an embodiment of the present application;

[0062] Figure 3 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 1 ;

[0063] Figure 4A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 2 ;

[0064] Figure 5 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 3 ;

[0065] Figure 6 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 4 ;

[0066] Figure 7 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 5 ;

[0067] Figure 8 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 6 ;

[0068] Fig. 9 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 7 ;

[0069] Fig.10 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 8 ;

[0070] Fig.11 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 9 ;

[0071] Fig.12 A structural schematic diagram of a crack detection device for a concrete dam provided in an embodiment of the present application. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0073] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0074] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0075] In the description of the present application, it should be noted that if the terms "upper", "lower", "inside", "outside", etc. appear to indicate an orientation or position relationship, it is based on the orientation or position relationship shown in the drawings, or is the orientation or position relationship in which the product of the invention is usually placed when used. It is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0076] In addition, the terms “first”, “second”, etc., if used, are merely used to distinguish between the descriptions and should not be understood as indicating or implying relative importance.

[0077] It should be noted that, in the absence of conflict, the features in the embodiments of the present application may be combined with each other.

[0078] In order to clearly describe a crack detection method for a concrete dam provided in an embodiment of the present application, a crack detection system for a concrete dam is first described in detail with reference to the accompanying drawings. Figure 1 This is a schematic diagram of the structure of a crack detection system for a concrete dam provided in an embodiment of the present application. Figure 1 As shown, the concrete dam crack detection system 100 may include: an image acquisition device 110 and a computer device 120 .

[0079] The image acquisition device 110 is connected to the computer device 120 for communication, and the communication connection mode may be wireless communication connection or wired communication connection.

[0080] The image acquisition device 110 may be integrated into a preset drone, so that the image acquisition device 110 integrated into the preset drone can acquire images of the concrete dam through the flight of the preset drone. The image acquisition device 110 may be selected according to actual conditions.

[0081] The computer device 120 can be an embedded edge computing device or a computing device with data encryption. There is no limitation here and it can be set according to actual conditions.

[0082] Among them, the computer device 120 can be connected to multiple image acquisition devices 110, and can simultaneously obtain multiple real-time images of concrete dams from multiple image acquisition devices 110, and then based on the multiple real-time images of concrete dams, perform detection and analysis on the concrete dam to be detected in the multiple real-time images of concrete dams.

[0083] The embodiment of the present application proposes a crack detection system for a concrete dam. Through the communication connection between the image acquisition device and the computer device, the real-time image of the concrete dam to be detected and the historical image of the concrete dam collected by the image acquisition device are transmitted to the computer device to perform crack detection and analysis in the real-time image of the concrete dam to be detected. Therefore, the fall detection system of the present application can realize the crack detection and analysis function in the real-time image of the concrete dam to be detected only through the image acquisition device and the computer device integrated on the preset drone, so that the crack detection system for the concrete dam provided by the embodiment of the present application is less complex, and the processing and calculation of the crack detection and analysis function in the real-time image of the concrete dam to be detected is less time-consuming, efficient and fast.

[0084] Furthermore, in order to clearly describe the concrete dam crack detection system 100 provided in the above embodiment, the present application also provides a structural schematic diagram of a computer device 120. Figure 2 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. Figure 2 As shown, the computer device 120 may include: a processor 121 and a memory 122 .

[0085] The memory 122 stores machine executable instructions that can be executed by the processor 121, that is, when the computer device 120 is running, the above machine readable instructions are executed, and the processor 121 communicates with the memory 122 via a bus. The processor 121 can execute the machine executable instructions to implement the crack detection method for concrete dams.

[0086] The memory 122, the processor 121 and the bus components are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines. The mobile storage device includes at least one software function module that can be stored in the memory 122 in the form of software or firmware or solidified in the operating system (OS) of the electronic device. The processor 121 is used to execute the executable modules stored in the memory 122, such as the software function modules and computer programs included in the crack detection method of the concrete dam of the mobile storage medium.

[0087] Among them, the memory 122 can be, but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.

[0088] The crack detection method for a concrete dam provided in the embodiment of the present application may be executed by the processor 121 in the computer device 120. The crack detection method for a concrete dam provided in the above embodiment of the present application will be explained and illustrated in detail below in conjunction with the accompanying drawings. Figure 3 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 1 .like Figure 3 As shown, the method may include:

[0089] S201. Acquire historical images of the concrete dam and real-time images of the concrete dam to be inspected.

[0090] Among them, the historical image of the concrete dam may include the original image when the concrete dam began to pass, the original depth image, the weather conditions after the concrete dam passed, and the number of days the concrete dam was passed, etc. The historical image of the concrete dam may be data reserved in a computer device or data previously collected by an image acquisition device carried by a preset drone.

[0091] In one possible implementation, the concrete dam will be damaged by weather conditions such as high temperature, low temperature, heavy weight, rain, etc. after the concrete dam passes through, but weather conditions such as low temperature, high temperature, rain, etc. are all normal damage. When the weight carried by the concrete dam is too large, the concrete dam is overloaded, causing crack damage to the concrete dam, but the manual inspection cycle is too long. Therefore, the real-time image of the concrete dam to be inspected is collected by the image acquisition device carried by the preset drone to inspect the cracks in the concrete dam to be inspected, which can not only shorten the crack inspection cycle, but also reduce the work intensity of the staff. At the same time, it is necessary to introduce weather conditions such as low temperature, high temperature, rain, etc. and normal load weight as reference standards for normal damage, thereby improving the detection accuracy of cracks in the concrete dam to be inspected.

[0092] It should be noted that the concrete dam historical images in the above embodiment may be in the same concrete dam project or in different concrete dam projects. That is to say, the concrete dam historical images may include multiple historical images of one concrete dam or historical images of multiple concrete dams. Moreover, the real-time image of the concrete dam to be detected may correspond to the above-mentioned concrete dam historical image, that is, the real-time image of the concrete dam to be detected may be the same as or different from the concrete dam historical image.

[0093] S202, performing denoising processing on the historical image and the real-time image of the concrete dam to obtain a denoised historical image and a denoised real-time image.

[0094] In one possible implementation, both the acquired historical image of the concrete dam and the real-time image of the concrete dam to be inspected are denoised to obtain a denoised historical image and a denoised real-time image, so as to ensure that the denoised historical image and the denoised real-time image are in the same standard to facilitate subsequent processing.

[0095] S203 , performing crack identification on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image.

[0096] In one possible implementation method, crack identification is performed on the denoised historical image and the denoised real-time image to obtain the historical crack area image and the real-time crack area image, so as to obtain data of the same standard, so that the historical crack area image and the real-time crack area image are of reference and researchable value.

[0097] S204, performing crack feature analysis on the historical crack region image and the real-time crack region image respectively, to obtain historical crack feature data corresponding to the historical crack region image and real-time crack feature data corresponding to the real-time crack region image.

[0098] In one possible implementation method, crack feature analysis is performed on the crack width, length and depth in the historical crack area image and the real-time crack area image respectively to obtain historical crack feature data corresponding to the historical crack area image and real-time crack feature data corresponding to the real-time crack area image.

[0099] S205: construct a crack recognition model based on the annotated dimension information corresponding to the historical crack characteristic data.

[0100] In a possible implementation, a crack recognition model is constructed based on preset machine learning according to the annotated dimension information corresponding to the historical crack feature data. The input data of the crack recognition model includes the historical crack feature data corresponding to the acquired historical crack region image of the concrete dam. The preset machine learning can be selected according to actual conditions, for example, the preset machine learning can be selected as a neural network model (such as a feedforward neural network model).

[0101] It should be noted that the feedforward neural network is a type of artificial neural network that adopts a unidirectional multi-layer structure, in which the 0th layer is called the input layer, the last layer is called the output layer, and the other intermediate layers are called hidden layers. The hidden layer can be one layer or multiple layers. Each hidden layer contains a number of neurons. Each neuron can receive signals from the neurons in the previous layer and generate outputs to the next layer. There is no feedback in the entire network, and the signal propagates unidirectionally from the input layer to the output layer.

[0102] S206: Input the real-time crack feature data into the crack identification model to obtain real-time crack size information.

[0103] In one possible implementation, the obtained real-time crack characteristic data is input into a crack identification model to obtain real-time crack size information, so that the staff can grasp the real-time crack size in real time.

[0104] S207 , performing crack detection on the concrete dam to be detected according to the real-time crack size information, and obtaining a real-time crack detection result of the concrete dam to be detected.

[0105] The real-time crack detection result is used to indicate whether the cracks in the concrete dam to be detected are dangerous cracks or normal cracks.

[0106] In one possible implementation, crack detection is performed on the concrete dam to be detected based on the real-time crack size information, and real-time crack detection results of the concrete dam to be detected can be obtained, so as to analyze and judge whether the cracks in the concrete dam to be detected are in danger or normal based on the real-time crack detection results, thereby realizing automatic detection of cracks in the concrete dam to be detected and shortening the detection cycle of cracks in the concrete dam.

[0107] The present application provides a method for detecting cracks in a concrete dam, which comprises acquiring a historical image of the concrete dam and a real-time image of the concrete dam to be detected; then denoising the historical image and the real-time image of the concrete dam to obtain a denoised historical image and a denoised real-time image; and performing crack recognition on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image, so as to denoise the historical image of the concrete dam and the real-time image of the concrete dam to be detected, analyze and correct the historical image of the concrete dam and the real-time image of the concrete dam to be detected, so that the data of the historical image of the concrete dam and the real-time image of the concrete dam to be detected are in the same standard, which is convenient for subsequent processing; the historical crack region image and the real-time crack region image are respectively Perform crack feature analysis to obtain historical crack feature data corresponding to the historical crack area image and real-time crack feature data corresponding to the real-time crack area image; then construct a crack recognition model based on the annotated size information corresponding to the historical crack feature data, that is, construct a crack recognition model that meets the requirements through a large amount of historical crack feature data; then input the real-time crack feature data into the crack recognition model to obtain real-time crack size information, so that the staff can grasp the real-time crack size information in real time; and perform crack detection on the concrete dam to be detected based on the real-time crack size information to obtain the real-time crack detection results of the concrete dam to be detected, so as to indicate whether the cracks of the concrete dam to be detected are dangerous cracks or normal cracks. Therefore, the present application can realize automatic detection of cracks in concrete dams, shorten the detection cycle of cracks in concrete dams, eliminate the need for staff to check cracks one by one, reduce safety hazards and the work intensity of staff, and at the same time, meet the needs of staff.

[0108] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Figure 4 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 2 .like Figure 4 As shown, in the above method, the historical image and the real-time image of the concrete dam are denoised to obtain the denoised historical image and the denoised real-time image, which may include:

[0109] S301, decomposing high-frequency information and low-frequency information of a historical image of a concrete dam to obtain historical high-frequency information and historical low-frequency information corresponding to the historical image of the concrete dam.

[0110] In one possible implementation, the high-frequency information of the historical image of the concrete dam is decomposed into first overlapping blocks of the same size by a preset first decomposition algorithm, and the first Euclidean distance between the first overlapping blocks is calculated to construct similar first overlapping blocks into a first group, thereby forming a first group domain, and then each first group is learned by using a preset second decomposition algorithm to form a first adaptive learning dictionary corresponding to each first group, thereby obtaining the historical high-frequency information corresponding to the historical image of the concrete dam.

[0111] A first local pixel grouping process is performed on the pixel points of the low-frequency information of the historical image of the concrete dam, and an unbiased estimation method with a preset error is used to approximately represent the similarity between the first local pixel block and the first target pixel block in the low-frequency information, so as to obtain a historical sample set of similar local pixel blocks, which is the historical low-frequency information corresponding to the historical image of the concrete dam.

[0112] The preset first decomposition algorithm can be selected according to actual conditions, for example, the preset first decomposition algorithm can be selected as a wavelet transform algorithm. The preset second decomposition algorithm can also be selected according to actual conditions, for example, the preset second decomposition algorithm can be selected as a singular value decomposition algorithm.

[0113] S302 , reconstructing high-frequency pixels and low-frequency pixels on historical high-frequency information and historical low-frequency information respectively, to obtain a reconstructed historical high-frequency component and a reconstructed historical low-frequency component.

[0114] In one possible implementation, based on the historical high-frequency information corresponding to the historical image of the concrete dam, a first sparse code corresponding to the historical high-frequency information is calculated by combining a preset third decomposition algorithm with a convex optimization problem, and the high-frequency pixels are reconstructed using the first sparse code and the first adaptive learning dictionary obtained in step S301 to obtain a reconstructed historical high-frequency component.

[0115] The historical low-frequency information (i.e., historical sample sets) is traversed pixel by pixel, and then denoised using a preset component analysis algorithm. The orthogonal transformation matrix is ​​obtained using a preset covariance matrix algorithm. The eigenvalue matrix is ​​then combined to remove the dimensions containing a small amount of information in each historical sample set to obtain the reconstructed historical low-frequency components.

[0116] The preset third decomposition algorithm can be selected according to actual conditions, for example, the preset third decomposition algorithm can be selected as the split Bregman algorithm (split Bregman iteration algorithm). The preset component analysis algorithm can be selected according to actual conditions, for example, the preset component analysis algorithm can be selected as the LPG-PCA (Principal Component Analysis) algorithm.

[0117] It should be noted that when the reconstructed historical high-frequency components and the reconstructed historical low-frequency components are aggregated and restored by a preset component analysis algorithm (such as the LPG-PCA algorithm), the first sparse code in the first group of domains in the reconstructed historical high-frequency components represents the natural image, and clearly and effectively characterizes the inherent local sparsity and non-local self-similarity of the natural image in a unified manner, and extracts the high-frequency part from the historical image of the concrete dam without being affected by noise. While effectively removing noise, the details of the historical image of the concrete dam are retained, achieving high sparsity and high restoration quality of the denoised historical image, and improving the visual performance, peak signal-to-noise ratio and structural similarity of the denoised historical image.

[0118] S303 , aggregating the reconstructed historical high-frequency components and the reconstructed historical low-frequency components to generate a denoised historical image.

[0119] In a possible implementation, the reconstructed historical high-frequency components and the reconstructed historical low-frequency components are aggregated by a preset wavelet inverse transform algorithm to obtain a denoised historical image, wherein the preset wavelet inverse transform algorithm can be selected according to actual conditions.

[0120] S304 , decomposing the high-frequency information and the low-frequency information of the real-time image to obtain the real-time high-frequency information and the real-time low-frequency information corresponding to the real-time image.

[0121] In one possible implementation, the high-frequency information of the real-time image of the concrete dam to be inspected is decomposed into second overlapping blocks of the same size by a preset first decomposition algorithm (such as a wavelet transform algorithm), and the second Euclidean distance between the second overlapping blocks is calculated to construct similar second overlapping blocks into a second group, thereby forming a second group domain, and then each second group is learned by using the preset second decomposition algorithm to form a second adaptive learning dictionary corresponding to each second group, thereby obtaining the high-frequency information corresponding to the real-time image of the concrete dam to be inspected.

[0122] Then, a second local pixel grouping process is performed on the pixel points of the low-frequency information of the real-time image of the concrete dam to be detected, and an unbiased estimation method with a preset error is used to approximately represent the similarity between the second local pixel block and the second target pixel block in the low-frequency information of the real-time image, so as to obtain a real-time sample set of similar local pixel blocks, which is the real-time low-frequency information corresponding to the real-time image of the concrete dam to be detected.

[0123] S305 , respectively reconstructing high-frequency pixels and low-frequency pixels on the real-time high-frequency information and the real-time low-frequency information to obtain a reconstructed real-time high-frequency component and a reconstructed real-time low-frequency component.

[0124] In one possible implementation, based on the high-frequency information corresponding to the real-time image of the concrete dam to be inspected, a second sparse code corresponding to the high-frequency information is calculated by combining a preset third decomposition algorithm (such as a split Bregman iterative algorithm) with a convex optimization problem, and the high-frequency pixels are reconstructed using the second sparse code and the second adaptive learning dictionary obtained in step S304 to obtain a reconstructed real-time high-frequency component.

[0125] The real-time low-frequency information (i.e., the real-time sample set) is traversed pixel by pixel, and then denoised using the preset component analysis algorithm. The orthogonal transformation matrix is ​​obtained through the preset covariance matrix algorithm, and then the eigenvalue matrix is ​​combined to remove the dimensions containing a small amount of information in each real-time sample set to obtain the reconstructed historical low-frequency components.

[0126] S306: Aggregate the reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component to generate a denoised real-time image.

[0127] In a possible implementation manner, the reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component are aggregated by a preset wavelet inverse transform algorithm to obtain a denoised real-time image.

[0128] Therefore, through the above steps, the denoised historical image and the denoised real-time image can be obtained.

[0129] It should be noted that in order to ensure that the denoised historical images and the denoised real-time images are at the same standard, both the historical images and the real-time images of the concrete dam need to be denoised according to the above steps, so that the denoised historical images and the denoised real-time images are of reference and research value.

[0130] The crack detection method of the concrete dam provided by the present application first decomposes the high-frequency information and low-frequency information of the historical image of the concrete dam to obtain the historical high-frequency information and historical low-frequency information corresponding to the historical image of the concrete dam, then reconstructs the historical high-frequency information and the historical low-frequency information by high-frequency pixels and low-frequency pixels respectively, obtains the reconstructed historical high-frequency component and the reconstructed historical low-frequency component, and then aggregates the reconstructed historical high-frequency component and the reconstructed historical low-frequency component to generate a denoised historical image; secondly, decomposes the high-frequency information and the low-frequency information of the real-time image to obtain the real-time high-frequency information and the real-time low-frequency information corresponding to the real-time image, then reconstructs the real-time high-frequency information and the real-time low-frequency information by high-frequency pixels and low-frequency pixels respectively, obtains the reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component; then aggregates the reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component to generate a denoised real-time image. Thus, the denoised historical image and the denoised real-time image can be obtained to ensure that the denoised historical image and the denoised real-time image are in the same standard, which is convenient for subsequent processing.

[0131] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Figure 5 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 3 .like Figure 5 As shown, in the above method, crack recognition is performed on the denoised historical image and the denoised real-time image to obtain the historical crack area image and the real-time crack area image, which may include:

[0132] S401 , respectively performing crack rectangular frame recognition on the denoised historical image and the denoised real-time image to obtain a historical crack rectangular frame and a real-time crack rectangular frame.

[0133] In one possible implementation method, in order to make the denoised historical image and the denoised real-time image at the same standard, the crack rectangular frame recognition is first performed on the denoised historical image and the denoised real-time image respectively to obtain the historical crack rectangular frame and the real-time crack rectangular frame, and then the cracks in the denoised historical image and the denoised real-time image are positioned according to the outer rectangle of the historical crack rectangular frame and the real-time crack rectangular frame.

[0134] S402 , respectively performing tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image to obtain an angle-corrected historical image and an angle-corrected real-time image.

[0135] In one possible implementation method, the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image are respectively corrected in tilt angle to obtain the angle-corrected historical image and the angle-corrected real-time image, and the tilt angle of the cracks in the denoised historical image and the denoised real-time image are initially corrected based on the angle-corrected historical image and the angle-corrected real-time image.

[0136] S403 , screening the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image to obtain a target historical rectangular frame and a target real-time rectangular frame.

[0137] S404 , according to the target historical rectangular frame and the target real-time rectangular frame, respectively determine the historical crack region image and the real-time crack region image from the denoised historical image and the denoised real-time image.

[0138] In one possible implementation method, the historical crack rectangular frame in the historical image after angle correction and the real-time crack rectangular frame in the real-time image after angle correction are screened to obtain the target historical rectangular frame and the target real-time rectangular frame, so as to determine the historical crack area image and the real-time crack area image from the denoised historical image and the denoised real-time image respectively, that is, to obtain the final corrected real-time crack area image and the final corrected historical crack area image, so that the final corrected real-time crack area image and the final corrected historical crack area image are in the same standard and have reference and researchability.

[0139] The crack detection method of the concrete dam provided by the present application firstly performs crack rectangular frame recognition on the denoised historical image and the denoised real-time image respectively to obtain the historical crack rectangular frame and the real-time crack rectangular frame; then performs tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image respectively to obtain the angle-corrected historical image and the angle-corrected real-time image; then screens the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image to obtain the target historical rectangular frame and the target real-time rectangular frame; finally, according to the target historical rectangular frame and the target real-time rectangular frame, determines the historical crack region image and the real-time crack region image from the denoised historical image and the denoised real-time image respectively. Thus, the present application recognizes the cracks appearing in the historical image and the real-time image of the concrete dam, and performs rectangular positioning and tilt angle correction on the cracks appearing, so that the historical image and the real-time image of the concrete dam are in the same standard.

[0140] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Figure 6 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 4 .like Figure 6 As shown, in the above method, crack rectangular frames are respectively identified on the denoised historical image and the denoised real-time image to obtain the historical crack rectangular frame and the real-time crack rectangular frame, which may include:

[0141] S501 , performing edge detection on the denoised historical image and the denoised real-time image respectively to obtain a historical crack edge and a real-time crack edge.

[0142] In one possible implementation, edge detection is performed on the denoised historical image and the denoised real-time image in the horizontal and vertical directions respectively, and the edge detection results in the horizontal and vertical directions are summed to obtain the historical crack edge and the real-time crack edge.

[0143] S502. According to the historical crack edges and the real-time crack edges, the denoised historical image and the denoised real-time image are processed respectively using a first morphological closing algorithm to obtain a historical path region image in the denoised historical image and a real-time path region image in the denoised real-time image.

[0144] In one possible implementation method, based on the historical crack edges and the real-time crack edges, the first morphological closing algorithm is used to process the parts containing cracks in the denoised historical image and the denoised real-time image, respectively, to obtain the historical path area image in the denoised historical image and the real-time path area image in the denoised real-time image.

[0145] Among them, the first morphological closing algorithm can be selected according to actual conditions.

[0146] S503: Using the second morphological closing algorithm, single line interference is removed from the historical path region image and the real-time path region image.

[0147] In one possible implementation, a second morphological closing algorithm is used to remove single lines in the historical passage area image and the real-time passage area image to avoid interference of the single lines in the historical passage area image and the real-time passage area image with the crack detection results.

[0148] Among them, the second morphological closing algorithm can be selected according to actual conditions.

[0149] S504: Fill holes in the historical path region image and the real-time path region image after interference removal.

[0150] In a possible implementation, holes are supplemented respectively on the historical path region image and the real-time path region image which are processed by the second morphological closing algorithm and are free of single line interference.

[0151] S505 , calculating the minimum enclosing rectangles according to each white area in the historical path region image after the hole filling and each white area in the real-time path region image, respectively, as the historical crack rectangular frame and the real-time crack rectangular frame.

[0152] In one possible implementation, calculations are performed separately based on each white area in the historical path area image after hole filling and each white area in the real-time path area image after hole filling. The following formula (1) can be used to calculate the minimum enclosing rectangle of each white area, and the historical crack rectangular frame and the real-time crack rectangular frame can be obtained.

[0153] The calculation formula of the minimum enclosing rectangle is:

[0154] S min =min(S1,S2,.....Sn ) Formula (1)

[0155] In the formula, S min is the minimum enclosing rectangle; S1,S2,.....S n is the area of ​​several groups of bounding rectangles.

[0156] Thus, each outer rectangle information, i.e., crack rectangle frame, can be obtained by calculating the minimum outer rectangle of each white area. Each outer rectangle information may include: coordinates of the four vertices of the outer rectangle, tilt angle, and number of white pixels in the outer rectangle area.

[0157] The crack detection method of the concrete dam provided by the present application first performs edge detection on the denoised historical image and the denoised real-time image respectively to obtain the historical crack edge and the real-time crack edge; then, according to the historical crack edge and the real-time crack edge, the first morphological closing algorithm is used to process the denoised historical image and the denoised real-time image respectively to obtain the historical passage area image in the denoised historical image and the real-time passage area image in the denoised real-time image; then, the second morphological closing algorithm is used to remove the single line interference of the historical passage area image and the real-time passage area image; and the holes are supplemented for the historical passage area image and the real-time passage area image after the interference is removed; finally, the minimum outer bounding rectangle is calculated according to each white area in the historical passage area image after the hole supplement and each white area in the real-time passage area image, as the historical crack rectangular frame and the real-time crack rectangular frame. Therefore, the present application can perform outer bounding rectangle positioning for the cracks in the denoised historical image and the denoised real-time image according to the historical crack rectangular frame and the real-time crack rectangular frame.

[0158] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Figure 7 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 5 .like Figure 7 As shown, in the above method, the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image are respectively corrected in tilt angles to obtain the angle-corrected historical image and the angle-corrected real-time image, which may include:

[0159] S601, according to the tilt angle of the historical crack rectangular frame and the tilt angle of the real-time crack rectangular frame, respectively counting the first angle distribution of the historical crack rectangular frame within a preset right angle range and the second angle distribution of the real-time crack rectangular frame within the preset right angle range.

[0160] In one possible implementation, based on the inclination angle of the historical crack rectangular frame and the inclination angle of the real-time crack rectangular frame, the first angle distribution of the historical crack rectangular frame within a preset right angle range and the second angle distribution of the real-time crack rectangular frame within the preset right angle range are respectively counted.

[0161] S602: Determine, from the first angle distribution and the second angle distribution, respectively, a first target tilt angle and a second target tilt angle whose angle distributions satisfy a preset distribution concentration condition.

[0162] The preset distribution concentration condition may be selected according to actual conditions. For example, the preset distribution concentration condition may be an inclination angle at which the distribution is relatively concentrated within a preset right-angle range.

[0163] In one possible implementation, a first target tilt angle and a second target tilt angle whose angle distributions satisfy a preset distribution concentration condition (eg, tilt angles with a relatively concentrated distribution in a preset right-angle range) are determined from the first angle distribution and the second angle distribution, respectively.

[0164] S603 , performing tilt angle correction on the historical crack rectangular frame and the real-time crack rectangular frame respectively according to the first target tilt angle and the second target tilt angle, to obtain an angle-corrected historical image and an angle-corrected real-time image.

[0165] In one possible implementation, preliminary tilt angle corrections are performed on the historical crack rectangular frame and the real-time crack rectangular frame respectively according to the first target tilt angle and the second target tilt angle to obtain a historical image after preliminary angle correction and a real-time image after angle correction.

[0166] The crack detection method for concrete dams provided by the present application, according to the tilt angle of the historical crack rectangular frame and the tilt angle of the real-time crack rectangular frame, respectively counts the first angle distribution of the historical crack rectangular frame within the preset right angle range and the second angle distribution of the real-time crack rectangular frame within the preset right angle range; then respectively determines the first target tilt angle and the second target tilt angle whose angle distribution meets the preset distribution concentration condition from the first angle distribution and the second angle distribution; finally, according to the first target tilt angle and the second target tilt angle, respectively corrects the tilt angle of the historical crack rectangular frame and the real-time crack rectangular frame, and obtains the angle-corrected historical image and the angle-corrected real-time image. Therefore, the present application can perform preliminary correction of the tilt angle of cracks in the denoised historical image and the denoised real-time image according to the angle-corrected historical image and the angle-corrected real-time image.

[0167] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Figure 8 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 6 .like Figure 8 As shown, in the above method, the historical crack rectangular frame in the historical image after angle correction and the real-time crack rectangular frame in the real-time image after angle correction are screened to obtain the target historical rectangular frame and the target real-time rectangular frame, which may include:

[0168] S701, determining historical candidate crack rectangular frames in the zero degree direction or the ninety degree direction and with errors within a preset angle range from all historical crack rectangular frames in the historical image after angle correction, and screening the historical candidate crack rectangular frames to obtain target historical rectangular frames.

[0169] The preset angle range may be selected according to actual conditions. For example, the preset angle range may be selected to be plus or minus ten degrees.

[0170] In one possible implementation method, historical candidate crack rectangular frames whose crack areas are located in the zero-degree direction or the ninety-degree direction and whose errors are within a preset angle range (such as plus or minus ten degrees) are determined from all historical crack rectangular frames in the historical image after preliminary angle correction, and rectangular screening is performed on the crack areas in the historical candidate crack rectangular frames in the zero-degree direction or the ninety-degree direction to determine the final positions and final angles of the historical candidate crack rectangular frames in the historical image. Then, the final positions and final angles of the historical candidate crack rectangular frames in the historical image are rotated one last time to obtain the final corrected historical crack area image, that is, the target historical rectangular frame.

[0171] S702, determining the real-time candidate crack rectangular frames in the zero degree direction or the ninety degree direction and with an error within a preset angle range from all the real-time crack rectangular frames in the historical image after angle correction, and screening the real-time candidate crack rectangular frames to obtain the target real-time rectangular frame.

[0172] In one possible implementation method, real-time candidate crack rectangular frames whose crack areas are located in the zero-degree direction or the ninety-degree direction and whose errors are within a preset angle range (such as plus or minus ten degrees) are determined from all real-time crack rectangular frames in the historical image after preliminary angle correction, and rectangular screening is performed on the crack areas in the real-time candidate crack rectangular frames in the zero-degree direction or the ninety-degree direction to determine the final positions and final angles of the historical candidate crack rectangular frames in the real-time image, and then the final positions and final angles of the historical candidate crack rectangular frames in the real-time image are rotated one last time to obtain the final corrected real-time crack area image, that is, the target real-time rectangular frame.

[0173] The crack detection method for concrete dams provided in the present application determines the historical candidate crack rectangular frames in the zero-degree direction or the ninety-degree direction and whose errors are within the preset angle range from all the historical crack rectangular frames in the historical image after angle correction, and screens the historical candidate crack rectangular frames to obtain the target historical rectangular frames; and determines the real-time candidate crack rectangular frames in the zero-degree direction or the ninety-degree direction and whose errors are within the preset angle range from all the real-time crack rectangular frames in the historical image after angle correction, and screens the real-time candidate crack rectangular frames to obtain the target real-time rectangular frames. Thus, the present application can obtain the final corrected real-time crack area image and the final corrected historical crack area image, so that the final corrected real-time crack area image and the final corrected historical crack area image are in the same standard, which is referenceable and researchable.

[0174] Optionally, the historical crack characteristic data may include: historical crack width characteristic data, historical crack length characteristic data, and historical crack depth data.

[0175] In a possible implementation, constructing a crack recognition model according to the annotated dimension information corresponding to the historical crack feature data may include:

[0176] A crack recognition model is constructed based on the annotated width dimension information corresponding to the historical crack width characteristic data, the annotated length dimension information corresponding to the historical crack length characteristic data, and the annotated depth dimension information corresponding to the historical crack depth data.

[0177] In one possible implementation method, based on a feedforward neural network model, the annotated width dimension information corresponding to the historical crack width feature data, the annotated length dimension information corresponding to the historical crack length feature data, and the annotated depth dimension information corresponding to the historical crack depth data are divided to obtain a training set, a validation set, and a test set corresponding to the historical crack feature data; then the training set, the validation set, and the test set are used to perform supervised training, validation, and testing on the crack recognition model to obtain a crack recognition model whose accuracy meets the preset accuracy requirements.

[0178] Among them, the preset accuracy rate can be selected according to actual conditions.

[0179] For example, multiple historical crack width feature data, multiple historical crack length feature data, and multiple historical crack depth data are feature-marked and divided as construction data, and then the construction data is randomly divided into a training set, a validation set, and a test set according to a preset ratio, and the divided training set, validation set, and test set are labeled. Among them, the preset ratio can be selected according to actual conditions, for example, the preset ratio can be selected as 8:1:1. Among them, the preset ratio can be selected as 8:1:1 as the preferred ratio of this application.

[0180] Among them, the training set is used to train the model, that is, to determine the model's weights and biases and other learning parameters; the validation set is used to verify each model after training multiple models in the training set, and record the model accuracy, and then select the best model and its corresponding parameters; the test set is only used once, that is, when evaluating the final model after the training is completed. It does not participate in the process of learning parameters or the process of selecting hyperparameters, but is only used for model evaluation. Based on the preset machine learning (such as the feedforward neural network model), the model is supervised and trained through the training set, the obtained model is verified through the validation set, and the trained model is evaluated through the test set. The parameters are continuously adjusted until the accuracy meets the preset accuracy requirements, and the crack recognition model can be output.

[0181] In the crack detection method for concrete dams provided by the present application, the historical crack feature data can be divided into: historical crack width feature data, historical crack length feature data, and historical crack depth data; then, according to the annotated dimension information corresponding to the historical crack feature data, a crack recognition model is constructed, including: constructing a crack recognition model according to the annotated width dimension information corresponding to the historical crack width feature data, the annotated length dimension information corresponding to the historical crack length feature data, and the annotated depth dimension information corresponding to the historical crack depth data. Thus, the present application can construct a crack recognition model whose accuracy meets the preset accuracy requirement based on the historical crack width feature data, the historical crack length feature data, and the historical crack depth data in the historical crack feature data.

[0182] Optionally, the real-time crack characteristic data may include: real-time crack width characteristic data, real-time crack length characteristic data, and real-time crack depth data. The real-time crack size information may include: real-time crack width size information, real-time crack length size information, and real-time crack depth size information.

[0183] In a possible implementation, inputting real-time crack feature data into a crack identification model to obtain real-time crack size information may include:

[0184] The real-time crack width characteristic data is input into the crack recognition model to obtain the real-time crack width size information.

[0185] In one possible implementation, real-time crack width feature data is input into a crack identification model that meets preset accuracy requirements to obtain real-time crack width dimension information.

[0186] The real-time crack length characteristic data is input into the crack recognition model to obtain the real-time crack length size information.

[0187] In one possible implementation, real-time crack length characteristic data is input into a crack identification model that meets preset accuracy requirements to obtain real-time crack length size information.

[0188] The real-time crack depth feature data is input into the crack recognition model to obtain the real-time crack depth size information.

[0189] In one possible implementation, real-time crack depth feature data is input into a crack identification model that meets preset accuracy requirements to obtain real-time crack depth size information.

[0190] Therefore, the present application can obtain real-time crack size information based on obtaining real-time crack width size information, real-time crack length size information and real-time crack depth size information.

[0191] The crack detection method for concrete dams provided by the present application, the real-time crack feature data includes: real-time crack width feature data, real-time crack length feature data, real-time crack depth data; the real-time crack size information includes: real-time crack width size information, real-time crack length size information and real-time crack depth size information; the real-time crack feature data is input into the crack recognition model to obtain the real-time crack size information, including: inputting the real-time crack width feature data into the crack recognition model to obtain the real-time crack width size information; inputting the real-time crack length feature data into the crack recognition model to obtain the real-time crack length size information; inputting the real-time crack depth feature data into the crack recognition model to obtain the real-time crack depth size information. Therefore, the present application can obtain the real-time crack size information based on the real-time crack width size information, the real-time crack length size information and the real-time crack depth size information, so that the staff can grasp the crack size in real time.

[0192] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Fig. 9 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 7 .like Fig. 9 As shown, in the above method, crack detection is performed on the concrete dam to be detected according to the real-time crack size information to obtain the real-time crack detection result of the concrete dam to be detected, which may include:

[0193] S801. Calculate real-time crack space size information based on real-time crack width dimension information, real-time crack length dimension information, and real-time crack depth dimension information.

[0194] In one possible implementation, width dimension information, length dimension information and depth dimension information of each real-time crack are obtained, and the real-time crack space size information is calculated based on the real-time crack width dimension information, the real-time crack length dimension information and the real-time crack depth dimension information.

[0195] S802: If the real-time crack space size information is greater than or equal to a preset crack space threshold, a first real-time crack detection result is generated to indicate that the crack of the concrete dam to be detected is a dangerous crack.

[0196] Among them, the preset crack space threshold can be selected according to actual conditions.

[0197] In one possible implementation, the real-time crack space size information is compared with a preset crack space threshold. If the real-time crack space size information is greater than or equal to the preset crack space threshold, a first real-time crack detection result is generated to indicate that the crack in the concrete dam to be detected is a dangerous crack, that is, the real-time crack space size information is greater than or equal to the preset crack space threshold, and the real-time crack detection result is a dangerous crack, then it needs to be re-poured and cured in time until the normal crack is restored.

[0198] S803: If the real-time crack space size information is smaller than a preset crack space threshold, a second real-time crack detection result is generated to indicate that the cracks in the concrete dam to be detected are normal cracks.

[0199] In a possible implementation, if the real-time crack space size information is less than the preset crack space threshold, a second real-time crack detection result is generated to indicate that the crack of the concrete dam to be detected is a normal crack. That is, the real-time crack space size information is less than the preset crack space threshold, and the real-time crack detection result is a normal crack, but the possibility of subsequent expansion is not excluded, so the present application also leads to continuing to process the normal crack.

[0200] The crack detection method for concrete dams provided by the present application calculates the real-time crack space size information according to the real-time crack width size information, the real-time crack length size information and the real-time crack depth size information; if the real-time crack space size information is greater than or equal to the preset crack space threshold, a first real-time crack detection result is generated to indicate that the cracks in the concrete dam to be detected are dangerous cracks, and it is necessary to recast and maintain them in time until the normal cracks are restored; if the real-time crack space size information is less than the preset crack space threshold, a second real-time crack detection result is generated to indicate that the cracks in the concrete dam to be detected are normal cracks. Therefore, the present application can realize the automatic detection of cracks in concrete dams and shorten the detection cycle of cracks in concrete dams.

[0201] The following is a detailed explanation and description of the crack detection method for a concrete dam provided in the above embodiment of the present application in conjunction with the accompanying drawings. Fig.10 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 8 .like Fig.10 As shown, the above method may also include:

[0202] S901. If the first target crack in the concrete dam to be detected is a normal crack, crack detection is performed on the concrete dam to be detected for a continuous preset time period to obtain first crack space size information of the first target crack after the preset time period.

[0203] The preset time period may be selected according to actual conditions. For example, the preset time period may be selected as 1 week.

[0204] In a possible implementation, since the concrete dam to be inspected with normal cracks may expand and change in the later stage, the present application also introduces the continuation of processing the normal cracks. That is, if the first target crack in the concrete dam to be inspected is a normal crack, the image acquisition of the preset drone is performed continuously for a preset time period (such as within 1 week), and the crack detection of the concrete dam to be inspected is performed by the above method to obtain the first crack space size information of the first target crack after the preset time period (such as within 1 week). Among them, the first crack space size information may include: width dimension information, length dimension information and depth dimension information of the first target crack.

[0205] S902: Calculate the crack expansion rate of the first target crack in the preset time period according to the second crack space size information of the first target crack before the preset time period.

[0206] In a possible implementation, the second crack space size information before a preset time period (such as within 1 week) is calculated based on the width size information, length size information, and depth size information of the first target crack. Then, the second crack space size information before the preset time period (such as within 1 week) and the first crack space size information after the preset time period (such as within 1 week) are subtracted and divided by the time difference before the preset time period (such as within 1 week) and after the preset time period (such as within 1 week), to obtain the crack expansion rate of the first target crack.

[0207] S903. Calculate the average crack expansion rate of the concrete dam to be inspected according to the crack expansion rates of all first target cracks.

[0208] In a possible implementation, all crack expansion rates of the first target crack are summed and divided by a preset number to obtain an average crack expansion rate of the concrete dam to be inspected, wherein the preset number refers to the number of all crack expansion rates of the first target crack.

[0209] S904: If the average crack expansion rate is greater than or equal to a preset crack expansion rate threshold, it is determined that the first target crack of the concrete dam to be detected is in an abnormal expansion state.

[0210] Among them, the preset crack expansion rate threshold can be selected according to actual conditions.

[0211] In one possible implementation, the difference between the average crack expansion rate and a preset crack expansion rate threshold is determined. If the average crack expansion rate is greater than or equal to the preset crack expansion rate threshold, it is determined that the first target crack of the concrete dam to be inspected is in an abnormal expansion state, and the crack at that location needs to be protected and reinforced.

[0212] S905: If the average crack expansion rate is less than a preset crack expansion rate threshold, it is determined that all first target cracks in the concrete dam to be inspected are in a normal expansion state.

[0213] In a possible implementation, if the average crack expansion rate is less than a preset crack expansion rate threshold, it is determined that all first target cracks in the concrete dam to be inspected are in a normal expansion state, and no measures may be taken.

[0214] It should be noted that if the first target crack in the concrete dam to be detected is an abnormal crack, the following measures need to be taken for the dangerous crack area: First, a mixing machine is used to mix the concrete raw materials according to a preset ratio; wherein the preset ratio can be selected according to the actual situation. The mixed concrete is poured into the dangerous crack area to maintain the poured concrete.

[0215] The crack detection method of the concrete dam provided by the present application, if the first target crack in the concrete dam to be detected is a normal crack, then the crack detection is performed on the concrete dam to be detected for a preset time period continuously, and the first crack space size information of the first target crack after the preset time period is obtained; according to the second crack space size information of the first target crack before the preset time period, the crack expansion rate of the first target crack in the preset time period is calculated; according to the crack expansion rate of all the first target cracks, the average crack expansion rate of the concrete dam to be detected is calculated; if the average crack expansion rate is greater than or equal to the preset crack expansion rate threshold, it is determined that the first target crack of the concrete dam to be detected is in an abnormal expansion state, and the crack at the location needs to be protected and reinforced; if the average crack expansion rate is less than the preset crack expansion rate threshold, it is determined that all the first target cracks of the concrete dam to be detected are in a normal expansion state, and no measures need to be taken. Therefore, the present application can realize the automatic detection of cracks in the concrete dam, shorten the detection cycle of cracks in the concrete dam, and reduce safety hazards and the work intensity of the staff.

[0216] To facilitate understanding of the above-mentioned method for detecting cracks in a concrete dam, the present application embodiment further provides an example of a process of the method for detecting cracks in a concrete dam, which will be described in conjunction with the accompanying drawings as follows: Fig.11 A flow chart of a method for detecting cracks in a concrete dam provided in an embodiment of the present application Figure 9 .like Fig.11 As shown, the crack detection method of the concrete dam may include:

[0217] S1001. Acquire historical images of the concrete dam and real-time images of the concrete dam to be inspected.

[0218] S1002: De-noising the historical image and the real-time image of the concrete dam to obtain a de-noised historical image and a de-noised real-time image.

[0219] S1003, identifying cracks in the denoised historical image and the denoised real-time image, and performing rectangular positioning and tilt angle correction on them, to obtain pre-processed historical images and real-time images of the concrete dam.

[0220] S1004. Based on the pre-processed historical image data and real-time image data of the concrete dam, feature analysis is performed on the width, length and depth of the cracks to obtain feature data of the width, length and depth of the cracks in the historical images and the real-time images.

[0221] S1005: construct a crack recognition model based on crack width feature data, length feature data and depth feature data in historical images.

[0222] S1006, inputting the crack width feature data, length feature data and depth feature data in the obtained real-time image into a crack recognition model, and outputting real-time crack information.

[0223] S1007: Analyze and process the real-time crack information to determine whether the real-time crack is normal.

[0224] By performing denoising analysis and correction processing on the historical image data and real-time image data of the concrete dam, the data are in the same standard, with reference and researchability. A crack recognition model that meets the preset requirements is constructed using a neural network model through a large amount of historical image data. The real-time image data is input into the crack recognition model to obtain real-time crack information, which is convenient for the staff to grasp the real-time crack information in real time. If the real-time crack is greater than or equal to the preset crack space threshold, it means that the real-time crack is more dangerous, and it needs to be re-cast and maintained in time until it returns to normal. If the real-time crack is less than the preset crack space threshold, it means that the real-time crack is normal, but the possibility of subsequent expansion is not ruled out. Therefore, the present application also leads to continuing to collect images of normal cracks, and calculating the average crack expansion rate within a preset time period (such as 1 week). If the average crack expansion rate is greater than or equal to the preset crack expansion rate threshold, it means that the real-time crack expansion speed is too large, and it needs to be reinforced and protected. Therefore, the crack detection method of the present application can be used to eliminate the need for workers to check cracks in concrete dams one by one, thereby reducing safety hazards for workers, speeding up crack detection in concrete dams, and meeting the workers' needs for crack detection in concrete dams.

[0225] The present application provides a method for detecting cracks in a concrete dam, which includes obtaining historical image data of the concrete dam in a database; using a drone to collect real-time images of the concrete dam to obtain real-time image data of the concrete dam; performing denoising processing on the historical image data and real-time image data of the concrete dam in turn to obtain preprocessing data of the historical image of the concrete dam and denoising processing data of the real-time image; identifying cracks appearing in the historical image and the real-time image of the concrete dam, and performing rectangular positioning and tilt angle correction on the cracks to obtain preprocessed historical images and real-time images of the concrete dam; performing feature analysis on the width, length and depth of the cracks based on the preprocessed historical image data and the real-time image data of the concrete dam to obtain crack width feature data, length feature data and depth feature data in the historical image and the real-time image; constructing a crack recognition model based on the crack width feature data, length feature data and depth feature data in the historical image; inputting the obtained crack width feature data, length feature data and depth feature data in the real-time image into the crack recognition model to output crack information; analyzing and processing the crack information to determine whether the crack is normal. Therefore, the present application can realize automatic detection of cracks in concrete dams, shorten the detection cycle of cracks in concrete dams, eliminate the need for workers to check cracks one by one, and reduce safety hazards and work intensity of workers.

[0226] Based on the same inventive concept, a crack detection device for a concrete dam is also provided in an embodiment of the present application. Since the principle of solving the problem by the device in the embodiment of the present application is similar to the crack detection method for a concrete dam in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0227] Fig.12 This is a schematic diagram of the structure of a crack detection device for a concrete dam provided in an embodiment of the present application. Fig.12 As shown, the concrete dam crack detection device 1100 may include:

[0228] An acquisition module 1101 is used to acquire historical images of concrete dams and real-time images of concrete dams to be inspected;

[0229] A denoising module 1102 is used to perform denoising on the historical image and the real-time image of the concrete dam to obtain a denoised historical image and a denoised real-time image;

[0230] The recognition module 1103 is used to perform crack recognition on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image;

[0231] An analysis module 1104 is used to perform crack feature analysis on the historical crack region image and the real-time crack region image respectively, to obtain historical crack feature data corresponding to the historical crack region image and real-time crack feature data corresponding to the real-time crack region image;

[0232] A construction module 1105 is used to construct a crack recognition model according to the annotated dimension information corresponding to the historical crack characteristic data;

[0233] An input module 1106 is used to input real-time crack feature data into a crack identification model to obtain real-time crack size information;

[0234] The detection module 1107 is used to perform crack detection on the concrete dam to be detected according to the real-time crack size information to obtain the real-time crack detection result of the concrete dam to be detected. The real-time crack detection result is used to indicate whether the crack of the concrete dam to be detected is a dangerous crack or a normal crack.

[0235] In an optional implementation, the denoising module 1102 is specifically used to: decompose the high-frequency information and low-frequency information of the historical image of the concrete dam to obtain the historical high-frequency information and historical low-frequency information corresponding to the historical image of the concrete dam; reconstruct the historical high-frequency information and the historical low-frequency information by high-frequency pixels and low-frequency pixels, respectively, to obtain the reconstructed historical high-frequency component and the reconstructed historical low-frequency component; aggregate the reconstructed historical high-frequency component and the reconstructed historical low-frequency component to generate a denoised historical image; decompose the high-frequency information and low-frequency information of the real-time image to obtain the real-time high-frequency information and real-time low-frequency information corresponding to the real-time image; reconstruct the real-time high-frequency information and the real-time low-frequency information by high-frequency pixels and low-frequency pixels, respectively, to obtain the reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component; aggregate the reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component to generate a denoised real-time image.

[0236] In an optional embodiment, the identification module 1103 is specifically used to: perform crack rectangular frame identification on the denoised historical image and the denoised real-time image, respectively, to obtain a historical crack rectangular frame and a real-time crack rectangular frame; perform tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image, respectively, to obtain an angle-corrected historical image and an angle-corrected real-time image; screen the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image, to obtain a target historical rectangular frame and a target real-time rectangular frame; determine the historical crack area image and the real-time crack area image from the denoised historical image and the denoised real-time image, respectively, according to the target historical rectangular frame and the target real-time rectangular frame.

[0237] In an optional embodiment, the identification module 1103 is specifically used to: perform edge detection on the denoised historical image and the denoised real-time image, respectively, to obtain historical crack edges and real-time crack edges; based on the historical crack edges and the real-time crack edges, use a first morphological closing algorithm to process the denoised historical image and the denoised real-time image, respectively, to obtain the historical path area image in the denoised historical image and the real-time path area image in the denoised real-time image; use a second morphological closing algorithm to remove single-line interference from the historical path area image and the real-time path area image; perform hole filling on the historical path area image and the real-time path area image after the interference is removed; and calculate the minimum enclosing rectangle based on each white area in the historical path area image after the hole filling and each white area in the real-time path area image, as the historical crack rectangular frame and the real-time crack rectangular frame.

[0238] In an optional embodiment, the identification module 1103 is specifically used to: according to the inclination angle of the historical crack rectangular frame and the inclination angle of the real-time crack rectangular frame, respectively count the first angle distribution of the historical crack rectangular frame within a preset right-angle range and the second angle distribution of the real-time crack rectangular frame within the preset right-angle range; respectively determine the first target inclination angle and the second target inclination angle whose angle distribution meets the preset distribution concentration condition from the first angle distribution and the second angle distribution; according to the first target inclination angle and the second target inclination angle, respectively correct the inclination angle of the historical crack rectangular frame and the real-time crack rectangular frame, to obtain the historical image after angle correction and the real-time image after angle correction.

[0239] In an optional embodiment, the identification module 1103 is specifically used to: determine the historical candidate crack rectangular frames that are in the zero degree direction or the ninety degree direction and whose errors are within a preset angle range from all historical crack rectangular frames in the historical image after angle correction, and screen the historical candidate crack rectangular frames to obtain the target historical rectangular frames; determine the real-time candidate crack rectangular frames that are in the zero degree direction or the ninety degree direction and whose errors are within a preset angle range from all real-time crack rectangular frames in the historical image after angle correction, and screen the real-time candidate crack rectangular frames to obtain the target real-time rectangular frames.

[0240] In an optional embodiment, the historical crack characteristic data includes: historical crack width characteristic data, historical crack length characteristic data, and historical crack depth data; the construction module 1105 is specifically used to: construct a crack recognition model based on the annotated width dimension information corresponding to the historical crack width characteristic data, the annotated length dimension information corresponding to the historical crack length characteristic data, and the annotated depth dimension information corresponding to the historical crack depth data.

[0241] In an optional embodiment, real-time crack characteristic data includes: real-time crack width characteristic data, real-time crack length characteristic data, real-time crack depth data; real-time crack size information includes: real-time crack width size information, real-time crack length size information and real-time crack depth size information; input module 1106 is specifically used to: input the real-time crack width characteristic data into the crack recognition model to obtain real-time crack width size information; input the real-time crack length characteristic data into the crack recognition model to obtain real-time crack length size information; input the real-time crack depth characteristic data into the crack recognition model to obtain real-time crack depth size information.

[0242] In an optional embodiment, the detection module 1107 is specifically used to: calculate the real-time crack space size information based on the real-time crack width dimension information, the real-time crack length dimension information and the real-time crack depth dimension information; if the real-time crack space size information is greater than or equal to a preset crack space threshold, generate a first real-time crack detection result to indicate that the cracks in the concrete dam to be detected are dangerous cracks; if the real-time crack space size information is less than the preset crack space threshold, generate a second real-time crack detection result to indicate that the cracks in the concrete dam to be detected are normal cracks.

[0243] In an optional embodiment, the detection module 1107 is further used for: if the first target crack in the concrete dam to be detected is a normal crack, then crack detection is performed on the concrete dam to be detected for a continuous preset time period to obtain the first crack space size information of the first target crack after the preset time period; based on the second crack space size information of the first target crack before the preset time period, the crack expansion rate of the first target crack in the preset time period is calculated; based on the crack expansion rates of all the first target cracks, the average crack expansion rate of the concrete dam to be detected is calculated; if the average crack expansion rate is greater than or equal to a preset crack expansion rate threshold, it is determined that the first target crack of the concrete dam to be detected is in an abnormal expansion state; if the average crack expansion rate is less than the preset crack expansion rate threshold, it is determined that all the first target cracks of the concrete dam to be detected are in a normal expansion state.

[0244] It should be noted that for details not disclosed in the crack detection device for a concrete dam in the embodiment of the present application, please refer to the details disclosed in the crack detection method for a concrete dam in the embodiment of the present application, and the details will not be repeated here.

[0245] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASICs), or one or more microprocessors, or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0246] Optionally, the embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the processor executes the steps of the crack detection method for a concrete dam using a mobile storage medium in the above embodiment. The specific implementation method and technical effect are similar and will not be repeated here.

[0247] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0248] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0249] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0250] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (English: Read-Only Memory, referred to as: ROM), a random access memory (English: Random Access Memory, referred to as: RAM), a disk or an optical disk, and other media that can store program codes. The above is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for detecting cracks in a concrete dam, characterized in that: include: Obtain historical images of concrete dams and real-time images of concrete dams to be inspected; De-noising the concrete dam historical image and the real-time image to obtain a de-noised historical image and a de-noised real-time image; Perform crack identification on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image; Performing crack feature analysis on the historical crack region image and the real-time crack region image respectively to obtain historical crack feature data corresponding to the historical crack region image and real-time crack feature data corresponding to the real-time crack region image; Constructing a crack recognition model according to the annotated dimension information corresponding to the historical crack characteristic data; Inputting the real-time crack characteristic data into the crack identification model to obtain real-time crack size information; The concrete dam to be detected is subjected to crack detection according to the real-time crack size information to obtain a real-time crack detection result of the concrete dam to be detected, wherein the real-time crack detection result is used to indicate whether the crack of the concrete dam to be detected is a dangerous crack or a normal crack.

2. The method for detecting cracks in a concrete dam according to claim 1, characterized in that: The denoising process is performed on the concrete dam historical image and the real-time image to obtain the denoised historical image and the denoised real-time image, including: Decomposing the high-frequency information and the low-frequency information of the historical image of the concrete dam to obtain the historical high-frequency information and the historical low-frequency information corresponding to the historical image of the concrete dam; Respectively reconstructing the historical high-frequency information and the historical low-frequency information by high-frequency pixels and low-frequency pixels to obtain a reconstructed historical high-frequency component and a reconstructed historical low-frequency component; Aggregating the reconstructed historical high-frequency component and the reconstructed historical low-frequency component to generate the denoised historical image; Decomposing the high-frequency information and the low-frequency information of the real-time image to obtain the real-time high-frequency information and the real-time low-frequency information corresponding to the real-time image; Respectively reconstructing high-frequency pixels and low-frequency pixels on the real-time high-frequency information and the real-time low-frequency information to obtain a reconstructed real-time high-frequency component and a reconstructed real-time low-frequency component; The reconstructed real-time high-frequency component and the reconstructed real-time low-frequency component are aggregated to generate the denoised real-time image.

3. The method for detecting cracks in a concrete dam according to claim 1, characterized in that: The performing crack identification on the denoised historical image and the denoised real-time image to obtain a historical crack region image and a real-time crack region image includes: Respectively performing crack rectangular frame recognition on the denoised historical image and the denoised real-time image to obtain a historical crack rectangular frame and a real-time crack rectangular frame; Performing tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image, respectively, to obtain an angle-corrected historical image and an angle-corrected real-time image; Screening the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image to obtain a target historical rectangular frame and a target real-time rectangular frame; According to the target historical rectangular frame and the target real-time rectangular frame, the historical crack region image and the real-time crack region image are determined from the denoised historical image and the denoised real-time image respectively.

4. The method for detecting cracks in a concrete dam according to claim 3, characterized in that: The step of performing crack rectangular frame recognition on the denoised historical image and the denoised real-time image respectively to obtain a historical crack rectangular frame and a real-time crack rectangular frame comprises: Performing edge detection on the denoised historical image and the denoised real-time image respectively to obtain a historical crack edge and a real-time crack edge; According to the historical crack edge and the real-time crack edge, the denoised historical image and the denoised real-time image are processed respectively using a first morphological closing algorithm to obtain a historical passage area image in the denoised historical image and a real-time passage area image in the denoised real-time image; Using the second morphological closing algorithm, single line interference removal is performed on the historical path area image and the real-time path area image; Filling holes in the historical path region image and the real-time path region image after interference removal respectively; The minimum enclosing rectangles are calculated according to each white area in the historical path region image after the hole supplementation and each white area in the real-time path region image, respectively, as the historical crack rectangular frame and the real-time crack rectangular frame.

5. The method for detecting cracks in a concrete dam according to claim 3, characterized in that: The step of respectively performing tilt angle correction on the historical crack rectangular frame in the denoised historical image and the real-time crack rectangular frame in the denoised real-time image to obtain an angle-corrected historical image and an angle-corrected real-time image includes: According to the inclination angle of the historical crack rectangular frame and the inclination angle of the real-time crack rectangular frame, respectively counting the first angle distribution of the historical crack rectangular frame within a preset right angle range and the second angle distribution of the real-time crack rectangular frame within the preset right angle range; Determine, from the first angle distribution and the second angle distribution, respectively, a first target tilt angle and a second target tilt angle whose angle distribution satisfies a preset distribution concentration condition; According to the first target tilt angle and the second target tilt angle, the tilt angles of the historical crack rectangular frame and the real-time crack rectangular frame are respectively corrected to obtain the angle-corrected historical image and the angle-corrected real-time image.

6. The method for detecting cracks in a concrete dam according to claim 3, characterized in that: The method of screening the historical crack rectangular frame in the angle-corrected historical image and the real-time crack rectangular frame in the angle-corrected real-time image to obtain a target historical rectangular frame and a target real-time rectangular frame includes: Determine, from all the historical crack rectangular frames in the angle-corrected historical image, the historical candidate crack rectangular frames that are in the zero-degree direction or the ninety-degree direction and whose errors are within a preset angle range, and screen the historical candidate crack rectangular frames to obtain the target historical rectangular frame; From all the real-time crack rectangular frames in the historical image after angle correction, determine the real-time candidate crack rectangular frames that are in the zero degree direction or the ninety degree direction and whose errors are within the preset angle range, and screen the real-time candidate crack rectangular frames to obtain the target real-time rectangular frame.

7. The method for detecting cracks in a concrete dam according to claim 1, characterized in that: The historical crack characteristic data includes: historical crack width characteristic data, historical crack length characteristic data, and historical crack depth data; The step of constructing a crack recognition model according to the annotated dimension information corresponding to the historical crack characteristic data includes: The crack recognition model is constructed according to the annotated width dimension information corresponding to the historical crack width characteristic data, the annotated length dimension information corresponding to the historical crack length characteristic data, and the annotated depth dimension information corresponding to the historical crack depth data.

8. The method for detecting cracks in a concrete dam according to claim 7, characterized in that: The real-time crack characteristic data includes: real-time crack width characteristic data, real-time crack length characteristic data, and real-time crack depth data; the real-time crack size information includes: real-time crack width size information, real-time crack length size information, and real-time crack depth size information; The step of inputting the real-time crack feature data into the crack identification model to obtain real-time crack size information includes: Inputting the real-time crack width characteristic data into the crack recognition model to obtain the real-time crack width dimension information; Inputting the real-time crack length characteristic data into the crack recognition model to obtain the real-time crack length size information; The real-time crack depth characteristic data is input into the crack recognition model to obtain the real-time crack depth dimension information.

9. The method for detecting cracks in a concrete dam according to claim 8, characterized in that: The step of performing crack detection on the concrete dam to be detected according to the real-time crack size information to obtain a real-time crack detection result of the concrete dam to be detected includes: Calculating real-time crack space size information according to the real-time crack width dimension information, the real-time crack length dimension information and the real-time crack depth dimension information; If the real-time crack space size information is greater than or equal to a preset crack space threshold, a first real-time crack detection result is generated to indicate that the crack of the concrete dam to be detected is a dangerous crack; If the real-time crack space size information is smaller than the preset crack space threshold, a second real-time crack detection result is generated to indicate that the crack of the concrete dam to be detected is a normal crack.

10. The method for detecting cracks in a concrete dam according to claim 9, characterized in that: The method further comprises: If the first target crack in the concrete dam to be detected is a normal crack, crack detection is performed on the concrete dam to be detected for a continuous preset time period to obtain first crack space size information of the first target crack after the preset time period; Calculating the crack expansion rate of the first target crack in the preset time period according to the second crack space size information of the first target crack before the preset time period; Calculating an average crack expansion rate of the concrete dam to be inspected according to the crack expansion rates of all the first target cracks; If the average crack expansion rate is greater than or equal to a preset crack expansion rate threshold, it is determined that the first target crack of the concrete dam to be inspected is in an abnormal expansion state; If the average crack expansion rate is less than the preset crack expansion rate threshold, it is determined that all the first target cracks in the concrete dam to be inspected are in a normal expansion state.

Images