Image recognition-based concrete surface defect detection method and system

Abstract

The application discloses a kind of concrete surface defect detection method and system based on image recognition, belong to image data processing technical field, by being preprocessed to the concrete surface image to be identified, obtain the concrete surface image to be identified after preprocessing, surface defect feature extraction algorithm is extracted to the concrete surface image to be identified after preprocessing with feature, obtain the image feature to be identified, finally, pre-deployed concrete defect identification model is called to the image feature to be identified with identification, obtains the surface defect detection result corresponding to each concrete surface image to be identified, not only can effectively detect concrete surface defect, but also can effectively improve defect detection precision, improve the concrete surface defect detection efficiency.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image data processing, and particularly relates to a concrete surface defect detection method and system based on image recognition. BACKGROUND

[0002] Concrete, commonly known as concrete, is a man-made stone material made by mixing cement, sand, gravel and water in a certain proportion. It has the advantages of high strength, good durability and strong plasticity, and is the main material for modern construction and infrastructure construction. Concrete can be made into various shapes and sizes according to needs, and is widely used in housing, bridges, roads, water conservancy and other engineering. After hardening, it forms a solid whole that can withstand large loads, and has good impermeability and frost resistance. With the development of science and technology, the types of concrete are constantly enriched, such as high-performance concrete, green concrete, etc., bringing more possibilities to the construction industry. As the main material of modern construction, the surface quality of concrete directly affects the safety and service life of the building. However, cracks, holes, pitted surfaces and other surface defects are prone to occur in concrete during pouring and curing. Traditional manual detection methods are low in efficiency and strong in subjectivity, and are difficult to meet the needs of large-scale and high-precision detection. SUMMARY

[0003] The application provides a concrete surface defect detection method and system based on image recognition to solve the problem of low efficiency in the existing concrete detection process.

[0004] In one aspect, the application provides a concrete surface defect detection method based on image recognition, comprising:

[0005] Determine the defect area to be identified, control the unmanned aerial vehicle to collect images of the defect area to be identified, and obtain at least one concrete surface image to be identified;

[0006] For any one of the concrete surface images to be identified, pre-process the concrete surface image to be identified to obtain the pre-processed concrete surface image to be identified;

[0007] Use a surface defect feature extraction algorithm to extract features from the pre-processed concrete surface image to be identified to obtain image features to be identified;

[0008] Call a pre-deployed concrete defect recognition model to identify the image features to be identified to obtain a surface defect detection result corresponding to each concrete surface image to be identified.

[0009] Further, after obtaining the surface defect detection result corresponding to each concrete surface image to be identified, further comprising:

[0010] According to the surface defect detection result corresponding to each to-be-identified concrete surface image, mark the area with defects on the map corresponding to the to-be-identified defect area.

[0011] Further, the to-be-identified concrete surface image is preprocessed to obtain the to-be-identified concrete surface image after preprocessing, including:

[0012] The to-be-identified concrete surface image is subjected to grayscale processing to obtain the to-be-identified concrete surface image after grayscale processing.

[0013] The to-be-identified concrete surface image after grayscale processing is subjected to grayscale nonlinear transformation to obtain the to-be-identified concrete surface image after grayscale nonlinear transformation.

[0014] The to-be-identified concrete surface image after grayscale nonlinear transformation is subjected to image sharpening processing to obtain the to-be-identified concrete surface image after preprocessing.

[0015] Further, the to-be-identified concrete surface image is subjected to grayscale processing to obtain the to-be-identified concrete surface image after grayscale processing, including:

[0016] The R value, the G value and the B value in the to-be-identified concrete surface image are determined to obtain the R value, the G value and the B value corresponding to each pixel point.

[0017] For any one pixel point in the to-be-identified concrete surface image, the maximum value among the R value, the G value and the B value is taken as the gray value corresponding to the pixel point to obtain the to-be-identified concrete surface image after grayscale processing.

[0018] Further, the to-be-identified concrete surface image after grayscale processing is subjected to grayscale nonlinear transformation to obtain the to-be-identified concrete surface image after grayscale nonlinear transformation, including: for any one pixel point in the to-be-identified concrete surface image after grayscale processing, the gray value of the pixel point is transformed by using gamma transformation to obtain the to-be-identified concrete surface image after grayscale nonlinear transformation.

[0019] Further, the to-be-identified concrete surface image after grayscale nonlinear transformation is subjected to image sharpening processing to obtain the to-be-identified concrete surface image after preprocessing, including: the to-be-identified concrete surface image after grayscale nonlinear transformation is subjected to image sharpening processing by using a Sobel operator to obtain the to-be-identified concrete surface image after preprocessing.

[0020] Further, a surface defect feature extraction algorithm is used to extract features from the to-be-identified concrete surface image after preprocessing to obtain to-be-identified image features, including:

[0021] The pre-processed concrete surface image to be identified is binarized to obtain a binarized concrete surface image to be identified.

[0022] The boundary feature of the binarized concrete surface image to be identified is extracted to obtain an image feature to be identified.

[0023] Further, the concrete defect identification model is set as a CNN model or a YOLO model.

[0024] Further, the pre-deployment method of the concrete defect identification model comprises: deploying the concrete defect identification model by using an improved pigeon swarm optimization algorithm.

[0025] In another aspect, the application provides a concrete surface defect detection system based on image recognition, comprising: a concrete image acquisition module, an image preprocessing module, an image feature extraction module, and a concrete defect detection module.

[0026] The concrete image acquisition module is used to determine a defect area to be identified, control a UAV to acquire images of the defect area to be identified, and acquire at least one concrete surface image to be identified.

[0027] The image preprocessing module is used to preprocess the concrete surface image to be identified for any one of the concrete surface images to be identified, and acquire a pre-processed concrete surface image to be identified.

[0028] The image feature extraction module is used to extract features of the pre-processed concrete surface image to be identified by using a surface defect feature extraction algorithm, and obtain an image feature to be identified.

[0029] The concrete defect detection module is used to identify the image feature to be identified by calling a pre-deployed concrete defect identification model, and acquire a surface defect detection result corresponding to each concrete surface image to be identified.

[0030] The concrete surface defect detection method and system based on image recognition provided by the application can effectively detect the concrete surface defects, effectively improve the defect detection accuracy, and improve the concrete surface defect detection efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0031] The accompanying drawings, which are incorporated herein and constitute part of the specification, illustrate embodiments consistent with the application and, together with the description, further serve to explain the principles of the application.

[0032] Figure 1 A flow chart of a concrete surface defect detection method based on image recognition provided by an embodiment of the application.

[0033] Figure 2 A structural schematic diagram of a concrete surface defect detection system based on image recognition provided by an embodiment of the application.

[0034] The specific embodiments of the application have been shown by the above-described drawings, and will be described in more detail hereinafter. These drawings and the written description are not intended to restrict the scope of the inventive concept in any way, but to illustrate the inventive concept to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0035] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The following description is made with reference to the accompanying drawings in which like reference numerals represent like elements, unless the context of use indicates otherwise. The following description of exemplary embodiments is not representative of all embodiments consistent with the present application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present application as detailed in the appended claims.

[0036] Embodiments of the application will be described in detail below with reference to the accompanying drawings.

[0037] As Figure 1 shown, an embodiment of the application provides a concrete surface defect detection method based on image recognition, comprising:

[0038] S101, determining a defect area to be identified, controlling a UAV to collect images of the defect area to be identified, and obtaining at least one concrete surface image to be identified;

[0039] The defect area to be identified can be a target area input by a worker through human-computer interaction, which contains coordinates in a specific world coordinate system (i.e., the target area contains corresponding latitude and longitude) to facilitate planning of an image collection route. An existing route planning algorithm can be used to plan a flight route of the UAV, so as to ensure that all positions of the defect area to be identified are collected.

[0040] S102, for any one of the concrete surface images to be identified, pre-processing the concrete surface image to be identified, and obtaining the concrete surface image to be identified after pre-processing;

[0041] The embodiment of the present application can effectively enhance the features related to defects of the image by pre-processing the concrete surface image to be identified, so as to improve the final defect recognition accuracy.

[0042] In S103, a surface defect feature extraction algorithm is used to extract features from the pre-processed concrete surface image to be identified, to obtain image features to be identified.

[0043] After pre-processing the concrete surface image to be identified, the features become easier to extract, so that the features of the pre-processed concrete surface image to be identified can be extracted to recognize the image features to be identified by using deep learning technology, so that the defect recognition can be realized.

[0044] In S104, a pre-deployed concrete defect recognition model is called to recognize the image features to be identified, to obtain the surface defect detection result corresponding to each concrete surface image to be identified.

[0045] The surface defect detection result can be the presence or absence of defects, or the specific defect type (such as cracks, holes, and pitted surface), which can be defined by the staff, so as to realize the self-defined detection task.

[0046] In the embodiment of the present application, after obtaining the surface defect detection result corresponding to each concrete surface image to be identified, the method further comprises:

[0047] According to the surface defect detection result corresponding to each concrete surface image to be identified, the area with defects is marked on the map corresponding to the defect area to be identified.

[0048] By marking the area with defects on the map corresponding to the defect area to be identified, a defect detection record can be formed, which is convenient for the staff to check, and can also be used as a basis for reminding the staff, so as to realize the automatic detection and notification of abnormalities.

[0049] In the embodiment of the present application, the pre-processing of the concrete surface image to be identified comprises:

[0050] The gray-scale processing is performed on the concrete surface image to be identified, to obtain the pre-processed concrete surface image to be identified.

[0051] The gray-scale nonlinear transformation is performed on the pre-processed concrete surface image to be identified, to obtain the pre-processed concrete surface image to be identified.

[0052] The image sharpening processing is performed on the pre-processed concrete surface image to be identified, to obtain the pre-processed concrete surface image to be identified.

[0053] The embodiment of the present application can effectively highlight the defect features, reduce the influence of other unnecessary features, and effectively improve the final recognition accuracy by sequentially performing gray processing, gray nonlinear transformation and image sharpening processing on the to-be-recognized concrete surface image.

[0054] In the embodiment of the present application, the to-be-recognized concrete surface image is subjected to gray processing to obtain a to-be-recognized concrete surface image after gray processing, including:

[0055] The R value, the G value and the B value in the to-be-recognized concrete surface image are determined to obtain the R value, the G value and the B value corresponding to each pixel point.

[0056] For any one pixel point in the to-be-recognized concrete surface image, the maximum value among the R value, the G value and the B value is taken as the gray value corresponding to the pixel point to obtain the to-be-recognized concrete surface image after gray processing.

[0057] In the embodiment of the present application, the to-be-recognized concrete surface image after gray processing is subjected to gray nonlinear transformation to obtain a to-be-recognized concrete surface image after gray nonlinear transformation, including: for any one pixel point in the to-be-recognized concrete surface image after gray processing, the gray value of the pixel point is transformed by using gamma transformation to obtain the to-be-recognized concrete surface image after gray nonlinear transformation.

[0058] In the embodiment of the present application, the to-be-recognized concrete surface image after gray nonlinear transformation is subjected to image sharpening processing to obtain a to-be-recognized concrete surface image after preprocessing, including: the to-be-recognized concrete surface image after gray nonlinear transformation is subjected to image sharpening processing by using a Sobel operator to obtain the to-be-recognized concrete surface image after preprocessing.

[0059] In the embodiment of the present application, a surface defect feature extraction algorithm is used to extract features from the to-be-recognized concrete surface image after preprocessing to obtain to-be-recognized image features, including:

[0060] The to-be-recognized concrete surface image after preprocessing is binarized to obtain a to-be-recognized concrete surface image after binarization.

[0061] For the to-be-recognized concrete surface image after binarization, boundary feature extraction is performed on the to-be-recognized concrete surface image to obtain to-be-recognized image features.

[0062] Any one pixel point, at least three adjacent pixel points exist around it, namely image corner point;And the pixel points on the image boundary exist at least five pixel points;The rest of the pixel points exist eight pixel points around;In order to improve the recognition effect of defects and reduce the influence of noise, in the embodiment of the application, for any one target pixel point, if the number of adjacent pixel points with value 1 around the target pixel point is less than 4, the value corresponding to the target pixel point is set to 0, otherwise it is set to 1, so as to further optimize the defect feature.

[0063] In the embodiment of the application, the concrete defect recognition model is set as a CNN (Convolutional Neural Network) model or a YOLO model. It is worth noting that in addition to the above-mentioned models, other image recognition models can also be used as the concrete defect recognition model.

[0064] In the embodiment of the application, the pre-deployment method of the concrete defect recognition model comprises: deploying the concrete defect recognition model by using an improved pigeon optimization algorithm.

[0065] Optionally, the improved pigeon optimization algorithm can comprise:

[0066] A1, the hyperparameters of the concrete defect recognition model are initialized to obtain a pigeon group;Wherein the pigeon group comprises a plurality of different pigeons, each pigeon contains part or all of the to-be-optimized parameters of the concrete defect recognition model, which can be customized by the staff.

[0067] For example, when the concrete defect recognition model is set as a CNN model, the pigeon can include all or part of the weights of the CNN model, so as to customize the training plan. In the process of initializing the hyperparameters, a random initialization method can be used for initialization.

[0068] A2, the loss function value corresponding to each pigeon is obtained by using the pre-acquired sample image features and the artificial defect labels corresponding to the sample image features;

[0069] The hyperparameters contained in any one pigeon can be applied to the concrete defect recognition model, and then the sample image features are taken as input and the artificial defect labels corresponding to the sample image features are taken as expected output to obtain the loss function value corresponding to each pigeon. The loss function value can be determined by the root mean square loss function or the cross entropy loss function.

[0070] A3, the flight process of the pigeon group is improved by using an adaptive inertia cooperation search strategy, and the pigeon group after one flight is obtained as:

[0071]

[0072]

[0073]

[0074] wherein, denotes the i-th pigeon in the i-th training process, t denotes the i-th pigeon after the first flight, i denotes the i-th pigeon after the first flight, denotes the i-th pigeon after the first flight, i denotes the total number of pigeons in the pigeon group, N denotes the i-th pigeon in the i-th training process, N denotes the i-th pigeon in the i-th training process, denotes the i-th pigeon in the i-th training process, t denotes the i-th pigeon in the i-th training process, i denotes the i-th pigeon in the i-th training process, denotes the i-th pigeon in the i-th training process, t denotes the i-th pigeon in the i-th training process, i denotes the i-th pigeon in the i-th training process, denotes a natural constant, denotes an inertia weight, R denotes a positioning track planning and a magnet, and R is a random number between 0 and 1, denotes a first random number between 0 and 1, denotes a second random number between 0 and 1, denotes the i-th pigeon in the i-th training process, t denotes the i-th pigeon in the i-th training process, i denotes the i-th pigeon in the i-th training process, denotes an optimal pigeon, i.e., a pigeon with the minimum loss function value; denotes a preset maximum value of the inertia weight, denotes a preset minimum value of the inertia weight, and T denotes the maximum number of training;

[0075] The embodiment of the application improves the flight process of the pigeon group by adopting the adaptive inertia cooperation search strategy, modifies the original flight to the optimal position to cooperative search flight, can effectively improve the local area search ability of the algorithm, and can effectively improve the local search ability of the algorithm. At the same time, the algorithm can have stronger global search ability and search speed in the early stage, and have more fine search ability in the later stage of the algorithm.

[0076] A4, the elite elimination strategy of the pigeon group is improved by adopting the mixed information learning strategy to obtain pigeons after the second flight, comprising:

[0077] The pigeons after the first flight are randomly paired two by two, and for any two paired pigeons, the pigeon with the smaller loss function value is taken as the winner, and the pigeon with the larger loss function value is taken as the loser;

[0078] the winner is updated as:

[0079]

[0080] in, Indicates the first t The m-th winner in the training process, where m = 1, 2, ..., N / 2. This represents the winner after the m-th second flight. This represents a third random number between (0,1). Represents the Cauchy distribution factor. This represents the smoothing transition factor, and log represents the sign of the logarithmic operation. This represents the fourth random number between (0,1). Indicates the Levi flight factor. This represents the average position of all the winners. Let G represent the fifth random number between (0,1), and let G represent the Gaussian distribution factor.

[0081] Update the loser to:

[0082]

[0083] in, Indicates the first t The nth winner in the training process, n=1,2,…,N / 2 This represents the loser after the nth second flight. Indicates the first learning factor. Indicates the second learning factor. This represents the sixth random number between (0,1). This represents the seventh random number between (0,1). Indicating the loser The corresponding winner, Indicating the loser The corresponding historical best value.

[0084] This invention employs a hybrid information learning strategy to improve the elite elimination strategy for pigeon flocks. It utilizes a competition-based speed-position update strategy to apply two different update strategies to the pigeons in the flock. Losers use the information of winners for updates, thereby improving the algorithm's convergence performance. Winners utilize Cauchy, Levy, and Gaussian distributions, ensuring their movement is not limited to small local areas, increasing population diversity and expanding the search range. As the algorithm progresses, the flock gradually converges, allowing the algorithm to maintain global search capabilities while improving convergence accuracy and ensuring normal execution.

[0085] The winners and losers of the second flight are both considered as pigeons after the second flight.

[0086] A5, if yes, determining the optimal pigeon according to the pigeon after the second flight, and taking the hyperparameters contained in the optimal pigeon as the final hyperparameters of the concrete defect identification model to realize the deployment of the concrete defect identification model, otherwise returning to step A2.

[0087] As shown in Figure 2 The embodiment of the application provides a concrete surface defect detection system based on image recognition, which comprises a concrete image acquisition module 201, an image preprocessing module 202, an image feature extraction module 203 and a concrete defect detection module 204.

[0088] The concrete image acquisition module 201 is used for determining a defect area to be recognized, controlling a UAV to acquire images of the defect area to be recognized, and acquiring at least one concrete surface image to be recognized.

[0089] The image preprocessing module 202 is used for preprocessing the concrete surface image to be recognized to obtain a preprocessed concrete surface image to be recognized.

[0090] The image feature extraction module 203 is used for extracting features of the preprocessed concrete surface image to be recognized by using a surface defect feature extraction algorithm to obtain image features to be recognized.

[0091] The concrete defect detection module 204 is used for calling a pre-deployed concrete defect identification model to recognize the image features to be recognized to obtain a surface defect detection result corresponding to each concrete surface image to be recognized.

[0092] The concrete surface defect detection system based on image recognition provided by the embodiment of the application can execute the above-mentioned method technical solution, and the principles and beneficial effects are similar, which will not be repeated here.

[0093] The concrete surface defect detection method and system based on image recognition provided by the embodiment of the application can preprocess the concrete surface image to be recognized to obtain a preprocessed concrete surface image to be recognized, extract features of the preprocessed concrete surface image to be recognized by using a surface defect feature extraction algorithm to obtain image features to be recognized, and finally call a pre-deployed concrete defect identification model to recognize the image features to be recognized to obtain a surface defect detection result corresponding to each concrete surface image to be recognized, which can effectively detect the concrete surface defect, effectively improve the defect detection precision and improve the concrete surface defect detection efficiency.

[0094] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer readable storage media (including, but not limited to, disk memory, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0095] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0096] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0097] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0098] Those skilled in the art will appreciate that all or portions of the operations and methods described can also be embodied in one or more programs that execute on one or more computers. A "program," as used in the description above, is software that can be read and executed by a computer. It is to be understood that the programs that implement methods can be stored on or transmitted across one or more computer readable media, such as any media that now exist or are hereafter developed including, but not limited to, RAM, ROM, EEPROM, CD-ROM or any other optical disk storage, magnetic disk storage or other memory storage devices.

[0099] The above detailed description of the specific embodiments of the present application has been given to understand the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An image recognition-based concrete surface defect detection method, characterized by, The method comprises the steps of: determining a to-be-identified defect area, controlling a UAV to collect images of the to-be-identified defect area, and obtaining at least one to-be-identified concrete surface image; for any one to-be-identified concrete surface image, pre-processing the to-be-identified concrete surface image to obtain a pre-processed to-be-identified concrete surface image; using a surface defect feature extraction algorithm to extract features from the pre-processed to-be-identified concrete surface image to obtain to-be-identified image features; calling a pre-deployed concrete defect identification model to identify the to-be-identified image features to obtain a surface defect detection result corresponding to each to-be-identified concrete surface image; the concrete defect identification model is a CNN model or a YOLO model; a pre-deployment method of the concrete defect identification model, comprising: A1, initializing the hyperparameters of the concrete defect identification model to obtain a pigeon group; A2, using pre-obtained sample image features and sample image feature corresponding artificial defect labels to obtain a loss function value corresponding to each pigeon; A3, using a self-adaptive inertia cooperation search strategy to improve the flight process of the pigeon group to obtain the pigeon group after one flight: wherein, denotes the i-th pigeon in the j-th training process, t denotes the i-th pigeon in the j-th training process, i denotes the i-th pigeon after the j-th flight, denotes the i-th pigeon after the j-th flight, i = 1, 2, …, i , N denotes the total number of pigeons in the pigeon group, N denotes the i-th pigeon in the j-th training process, denotes the i-th pigeon in the j-th training process, t denotes the i-th pigeon in the j-th training process, i denotes the i-th pigeon in the j-th training process, denotes the i-th pigeon in the j-th training process, t denotes the i-th pigeon in the j-th training process, i denotes the i-th pigeon in the j-th training process, denotes a natural constant, denotes an inertia weight, R denotes a positioning track planning and a magnet, and R is a random number between 0 and 1, denotes a first random number between 0 and 1, denotes a second random number between 0 and 1, denotes the i-th pigeon in the j-th training process, t denotes the i-th pigeon in the j-th training process, i denotes the i-th pigeon in the j-th training process, denotes an optimal pigeon, i.e., a pigeon with the minimum loss function value; denotes a preset maximum value of the inertia weight, denotes a preset minimum value of the inertia weight, and T denotes a maximum training number. A4, using a hybrid information learning strategy to improve the elite elimination strategy of the pigeon group to obtain the pigeons after two flights, comprising: randomly pairing the pigeons after one flight two by two, and for any two paired pigeons, the pigeon with a smaller loss function value is regarded as the winner, and the pigeon with a larger loss function value is regarded as the loser; updating the winner as: wherein, denotes the mth winner in the first training process, m = 1, 2, …, N / 2, t denotes the mth winner in the second training process, m = 1, 2, …, N / 2, denotes the winner after the mth second flight, denotes a third random number between (0, 1), denotes a Cauchy distribution factor, denotes a smooth transition factor, and log denotes a symbol of logarithmic operation, denotes a fourth random number between (0, 1), denotes a Levy flight factor, denotes an average position of all winners, denotes a fifth random number between (0, 1), and G denotes a Gaussian distribution factor. updating the loser as: wherein, denotes the n-th winner in the first training process, n = 1, 2,..., N / 2, t denotes the n-th winner in the second training process, n = 1, 2,..., N / 2, denotes the n-th loser after the second flight, n = 1, 2,..., N / 2, denotes the first learning factor, denotes the second learning factor, denotes the sixth random number between (0, 1), denotes the seventh random number between (0, 1), denotes the loser the corresponding winner, denotes the loser the corresponding historical optimal value; A5, determining whether the current training number is greater than or equal to the preset maximum training number, if yes, determining the optimal pigeon again according to the pigeons after two flights, and regarding the hyperparameters contained in the optimal pigeon as the final hyperparameters of the concrete defect identification model to realize the deployment of the concrete defect identification model, otherwise returning to step A2.

2. The image recognition-based concrete surface defect detection method of claim 1, wherein, After obtaining the surface defect detection result corresponding to each to-be-identified concrete surface image, the method further comprises: labeling the area with defects on a map corresponding to the to-be-identified defect area according to the surface defect detection result corresponding to each to-be-identified concrete surface image.

3. The image recognition-based concrete surface defect detection method of claim 1, wherein, The pre-processing of the to-be-identified concrete surface image to obtain the pre-processed to-be-identified concrete surface image comprises: gray processing the to-be-identified concrete surface image to obtain a gray-processed to-be-identified concrete surface image; performing gray nonlinear transformation on the gray-processed to-be-identified concrete surface image to obtain a gray nonlinearly transformed to-be-identified concrete surface image; performing image sharpening processing on the gray nonlinearly transformed to-be-identified concrete surface image to obtain the pre-processed to-be-identified concrete surface image.

4. The image recognition-based concrete surface defect detection method according to claim 3, characterized in that, The gray processing of the to-be-identified concrete surface image to obtain the gray-processed to-be-identified concrete surface image comprises: determining R, G and B values in the to-be-identified concrete surface image to obtain R, G and B values corresponding to each pixel point; For any pixel point in the to-be-recognized concrete surface image, the maximum value of the R value, the G value and the B value is taken as a gray value corresponding to the pixel point, to obtain a to-be-recognized concrete surface image after gray-scale processing.

5. The image recognition-based concrete surface defect detection method of claim 4, wherein, The to-be-recognized concrete surface image after the gray-scale processing is subjected to gray-scale nonlinear transformation, to obtain a to-be-recognized concrete surface image after the gray-scale nonlinear transformation, including: for any pixel point in the to-be-recognized concrete surface image after the gray-scale processing, the gray value of the pixel point is subjected to transformation by using gamma transformation, to obtain the to-be-recognized concrete surface image after the gray-scale nonlinear transformation.

6. The image recognition-based concrete surface defect detection method of claim 5, wherein, The to-be-recognized concrete surface image after the gray-scale nonlinear transformation is subjected to image sharpening processing, to obtain a to-be-recognized concrete surface image after preprocessing, including: the to-be-recognized concrete surface image after the gray-scale nonlinear transformation is subjected to image sharpening processing by using a Sobel operator, to obtain the to-be-recognized concrete surface image after the preprocessing.

7. The image recognition-based concrete surface defect detection method of claim 1, wherein, The to-be-recognized concrete surface image after the preprocessing is subjected to feature extraction by using a surface defect feature extraction algorithm, to obtain to-be-recognized image features, including: The to-be-recognized concrete surface image after the preprocessing is subjected to binarization, to obtain a to-be-recognized concrete surface image after binarization; For the to-be-recognized concrete surface image after the binarization, the to-be-recognized concrete surface image is subjected to boundary feature extraction, to obtain the to-be-recognized image features.

8. An image recognition-based concrete surface defect detection system for performing the image recognition-based concrete surface defect detection method according to any one of claims 1 to 7, characterized by, including: a concrete image acquisition module, an image preprocessing module, an image feature extraction module and a concrete defect detection module; The concrete image acquisition module is configured to determine a to-be-recognized defect region, control a UAV to acquire images of the to-be-recognized defect region, and acquire at least one to-be-recognized concrete surface image. The image preprocessing module is configured to, for any to-be-recognized concrete surface image, preprocess the to-be-recognized concrete surface image, and acquire a to-be-recognized concrete surface image after the preprocessing. The image feature extraction module is configured to extract features of the to-be-recognized concrete surface image after the preprocessing by using a surface defect feature extraction algorithm, to obtain to-be-recognized image features. The concrete defect detection module is configured to call a pre-deployed concrete defect recognition model to recognize the to-be-recognized image features, and acquire a surface defect detection result corresponding to each to-be-recognized concrete surface image.

Images