A dam crack detection system based on unmanned aerial vehicle technology

Abstract

The application belongs to the field of dam crack detection, and discloses a dam crack detection system based on unmanned aerial vehicle technology, which comprises an unmanned aerial vehicle and an image processing terminal; the unmanned aerial vehicle is used for shooting the slope of the dam to obtain multiple slope images; the image processing terminal is used for identifying the slope images to obtain a detection result; wherein, the shooting of the slope of the dam comprises: obtaining multiple shooting heights based on the height of the slope of the dam; the unmanned aerial vehicle moves horizontally at each shooting height, and keeps the distance between the unmanned aerial vehicle and the slope of the dam in the horizontal direction unchanged during the movement; the slope of the dam is shot once every fixed moving distance to obtain a slope image. The application greatly improves the efficiency of detecting the cracks of the slope of the dam.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of dam crack detection, and particularly relates to a dam crack detection system based on unmanned aerial vehicle technology. BACKGROUND

[0002] Dam crack detection refers to monitoring and evaluating the dam structure to identify and analyze cracks and other potential defects. This process is crucial to ensure the safety and long-term stability of the dam, especially under water pressure and other environmental factors.

[0003] The traditional crack detection method for the inclined surface of the dam is visual inspection. The visual inspection method has the problem of low detection efficiency, and therefore needs to be improved. SUMMARY

[0004] The purpose of the present application is to disclose a dam crack detection system based on unmanned aerial vehicle technology, which solves the technical problems proposed in the background art.

[0005] In order to achieve the above purpose, the present application provides the following technical scheme:

[0006] The present application provides a dam crack detection system based on unmanned aerial vehicle technology, comprising an unmanned aerial vehicle and an image processing terminal;

[0007] The unmanned aerial vehicle is used to take pictures of the inclined surface of the dam to obtain multiple inclined surface images;

[0008] The image processing terminal is used to identify the inclined surface images to obtain a detection result;

[0009] Wherein, the taking pictures of the inclined surface of the dam comprises:

[0010] Obtaining multiple shooting heights based on the height of the inclined surface of the dam;

[0011] The unmanned aerial vehicle moves horizontally at each shooting height, and maintains the same distance between the inclined surface of the dam in the horizontal direction during the movement. The inclined surface of the dam is photographed once every fixed movement distance to obtain an inclined surface image.

[0012] Preferably, the image processing terminal is a computer, which can establish a connection with the unmanned aerial vehicle through a WiFi connection or a data line connection. After the connection is established, the computer can access the stored inclined surface images in the storage device of the unmanned aerial vehicle.

[0013] Preferably, the multiple shooting heights are obtained based on the height of the inclined surface of the dam, comprising:

[0014] Let H represent the height of the inclined surface of the dam;

[0015] D represents the distance between the UAV and the slope of the dam in the horizontal direction when the slope is photographed;

[0016] N represents the number of photographing heights, N can make the slope images obtained at adjacent photographing heights have an overlapping area in the vertical direction when the UAV photographs the slope of the dam based on D;

[0017] The first photographing height is

[0018] The nth photographing height is

[0019] Preferably, the relationship between the area of the overlapping area and the area of the slope image is:

[0020]

[0021] Sa represents the area of the overlapping area in the slope image, Sb represents the area of the slope image, and h represents the set proportion.

[0022] Preferably, the value of h is 0.5.

[0023] Preferably, the slope image is identified to obtain a detection result, including:

[0024] Each slope image is preprocessed to obtain a preprocessed slope image;

[0025] The preprocessed slope images are spliced to obtain a spliced image;

[0026] The spliced image is identified using a detection algorithm based on deep learning to obtain a detection result.

[0027] Preferably, the slope image is preprocessed, including:

[0028] The slope image is subjected to distortion correction processing to obtain a distortion corrected image;

[0029] The distortion corrected image is subjected to illumination correction processing to obtain an illumination corrected image;

[0030] The illumination corrected image is subjected to noise reduction processing to obtain a preprocessed slope image.

[0031] Preferably, the preprocessed slope images are spliced to obtain a spliced image, including:

[0032] The following algorithm is used for image splicing to obtain a spliced image:

[0033] Feature point detection is performed on two images p1 and p2 to be spliced to obtain feature points in p1 and p2;

[0034] acquire the same feature points in p1 and p2 using a matching algorithm to obtain a plurality of feature point pairs;

[0035] remove the incorrect feature point pairs in the plurality of feature point pairs to obtain feature point pairs for image registration;

[0036] estimate the geometric transformation matrix between p1 and p2 based on the feature point pairs for image registration;

[0037] transform p1 into the coordinate system in which p2 is located based on the geometric transformation matrix to obtain image p3, and perform weighted average calculation on the gray values of the pixel points in p2 and p3 in the coordinate system in which p2 is located to obtain the gray value of the pixel point in the spliced image.

[0038] Preferably, the matching algorithm includes any one of FLANN and KNN.

[0039] Preferably, removing the incorrect feature point pairs in the plurality of feature point pairs to obtain the feature point pairs for image registration includes:

[0040] using the RANSAC algorithm to remove the incorrect feature point pairs in the plurality of feature point pairs to obtain the feature point pairs for image registration.

[0041] Beneficial effects:

[0042] Compared with the existing crack detection method, the plurality of shooting heights are acquired according to the height of the dam, then the slope of the dam is shot by moving along the horizontal direction at each shooting height using the unmanned aerial vehicle, the slope image of the slope of the dam is collected, then the spliced image obtained based on the slope image splicing is identified using the detection algorithm based on deep learning, thereby greatly improving the efficiency of detecting the cracks of the slope of the dam. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0044] Figure 1 is a schematic diagram of a dam crack detection system based on unmanned aerial vehicle technology.

[0045] Figure 2 is a schematic diagram of the process of pre-processing the slope image. DETAILED DESCRIPTION

[0046] Clearly, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0047] As shown in Figure 1 The present application provides a dam crack detection system based on unmanned aerial vehicle technology, comprising an unmanned aerial vehicle and an image processing terminal;

[0048] The unmanned aerial vehicle is used for shooting the slope of the dam to obtain multiple slope images;

[0049] The image processing terminal is used for identifying the slope images to obtain a detection result;

[0050] Wherein, the shooting of the slope of the dam comprises:

[0051] Based on the height of the slope of the dam, multiple shooting heights are obtained;

[0052] The unmanned aerial vehicle moves horizontally at each shooting height, and keeps the same distance between the slope of the dam in the horizontal direction during the movement. Every fixed moving distance, the slope of the dam is shot to obtain a slope image.

[0053] The present application obtains multiple shooting heights according to the height of the dam, and then uses the unmanned aerial vehicle to move along the horizontal direction at each shooting height to shoot the slope of the dam, realizes the collection of the slope image of the slope of the dam, and then uses the detection algorithm based on deep learning to identify the splicing image obtained based on the slope image splicing, thereby greatly improving the efficiency of detecting the cracks of the slope of the dam.

[0054] Preferably, the image processing terminal is a computer, which can establish a connection with the unmanned aerial vehicle through a WiFi connection or a data line connection. After the connection is established, the computer can access the shot slope images stored in the storage device of the unmanned aerial vehicle.

[0055] Specifically, the unmanned aerial vehicle can be controlled by the staff in charge of detection, or can be automatically flown after the flight route is planned.

[0056] Preferably, based on the height of the slope of the dam, multiple shooting heights are obtained, comprising:

[0057] H represents the height of the slope of the dam;

[0058] D represents the distance between the unmanned aerial vehicle and the slope of the dam in the horizontal direction during shooting.

[0059] N represents the number of shooting heights, and N is capable of causing the adjacent shooting height images to have an overlapping area in the vertical direction when the UAV shoots the slope of the dam based on D;

[0060] The first shooting height is

[0061] The nth shooting height is

[0062] Specifically, D can be obtained as follows:

[0063] Obtain a horizontal plane Q in which the UAV is located;

[0064] Obtain a curve S formed by the intersection of the horizontal plane Q and the slope of the dam;

[0065] Connect a point node of the curve S and the UAV to obtain a line segment Snode, the line segment Snode being perpendicular to the moving direction of the UAV; the length of the line segment Snode is D.

[0066] Preferably, the fixed moving distance is capable of causing the two adjacent slope images obtained at the same shooting height to have an overlapping area in the horizontal direction.

[0067] Preferably, the relationship between the area of the overlapping area and the area of the slope image is:

[0068]

[0069] Sa represents the area of the overlapping area in the slope image, Sb represents the area of the slope image, and h represents a set proportion.

[0070] Preferably, the value of h is 0.5.

[0071] Preferably, the slope image is identified to obtain a detection result, including:

[0072] Each slope image is preprocessed to obtain a preprocessed slope image;

[0073] The preprocessed slope images are spliced to obtain a spliced image;

[0074] The spliced image is identified using a deep learning-based detection algorithm to obtain a detection result.

[0075] Specifically, the deep learning-based detection algorithm can be a YOLOv8 algorithm.

[0076] The training process of the YOLOv8 algorithm is as follows:

[0077] 1. Prepare the dataset:

[0078] Data collection:

[0079] To use YOLOv8 to detect cracks in dam slopes, first, you need to collect a large number of labeled images, ensuring that the images contain different types, angles, lighting conditions, sizes, and other situations of cracks.

[0080] Image sources: You can use images of the dam surface taken by drones or use on-site monitoring pictures of the dam.

[0081] Image quality requirements: The images should be as clear as possible, containing the characteristics of cracks, and ensuring that the cracks are sufficiently prominent in the images.

[0082] Data diversity: In order to make the model have better generalization ability, the dataset should contain images of various backgrounds, lighting, weather conditions, and different angles.

[0083] Data annotation:

[0084] Use annotation tools to frame and label the cracks in the images. Common annotation tools include LabelImg, LabelMe, etc.

[0085] Annotation format: YOLO requires annotation files to be stored in.txt format, with each image corresponding to a text file. Each line in the file represents the class and coordinates of a crack box.

[0086] For each target (crack), the annotation format of YOLOv8 is:

[0087] <class_id><center_x><center_y> <width> <height>

[0088] class_id: ID of the crack class, assuming there is only one class of cracks with ID 0.

[0089] center_x and center_y: coordinates of the center point of the crack bounding box, normalized to the range [0, 1].

[0090] width and height: width and height of the crack bounding box, also normalized to the range [0, 1].

[0091] For example, to annotate the position of a crack in an image as follows: 0 0.5 0.5 0.2 0.1

[0093] This represents a crack with ID 0, whose center point coordinates are the center of the image, width is 0.2, and height is 0.1.

[0094] Data division:

[0095] Divide the dataset into training, validation, and test sets, usually in the following proportions:

[0096] Training set: 70-80% (used for model training);

[0097] Validation set: 10-15% (used for model tuning);

[0098] Test set: 10-15% (used to evaluate model performance);

[0099] 2. Install YOLOv8 environment:

[0100] Install dependencies:

[0101] YOLOv8 can be trained and inferred through the official library provided by Ultralytics. First, install the relevant Python packages and dependencies:

[0102] pip install ultralytics;

[0103] 3. Configure training parameters

[0104] Data configuration

[0105] Before starting training, you need to set up the configuration file required by YOLOv8. Create a.yaml file to define the structure of the dataset. This configuration file contains class labels and data paths. Assume there is a configuration file named crack_detection.yaml with the following content:

[0106] yamltrain: / path / to / train / images;

[0107] val: / path / to / val / images;

[0108] nc:1# Number of categories, cracks have only one category

[0109] names:['crack']#Category Name

[0110] train: The path to the training images.

[0111] val: The path to the verification image.

[0112] nc: Number of categories, set to 1 here because this invention only identifies cracks.

[0113] names: Category name; here there is only one category, "crack".

[0114] Configuration file (YAML) location:

[0115] The crack_detection.yaml configuration file needs to be placed in the training directory; the path can be set according to the actual situation.

[0116] 4. Select a pre-trained model:

[0117] YOLOv8 provides several pre-trained models, allowing you to choose a suitable model as a base for transfer learning. You can start training with a model provided in the official YOLOv8 GitHub repository. The choice of pre-trained model depends on factors such as the complexity of the training data and hardware resources.

[0118] Pre-trained model download:

[0119] YOLOv8 offers pre-trained models of different sizes (such as YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, YOLOv8x, etc.), allowing you to choose the appropriate pre-trained model according to your needs.

[0120] 5. Start training the model:

[0121] Start the YOLOv8 training process using the following command:

[0122] yolo task=detect mode=train model=yolov8n.pt data=crack_detection.yaml epochs=50batch=16imgsz=640

[0123] Command parameters explanation:

[0124] task=detect: Specifies the task as object detection.

[0125] mode=train: Indicates that training is performed.

[0126] model=yolov8n.pt: Selects a pre-trained YOLOv8 model, yolov8n.pt is a smaller model, you can also choose larger models such as yolov8l.pt, according to hardware performance and accuracy requirements.

[0127] data=crack_detection.yaml: Specifies the data set configuration file.

[0128] epochs=50: Number of training rounds (can be adjusted according to the size of the data set and training results).

[0129] batch=16: Batch size for each iteration (adjust according to GPU memory).

[0130] imgsz=640: Size of input image.

[0131] 6. Monitor the training process:

[0132] During training, YOLOv8 will output the loss function, precision, recall rate and other indicators of each epoch. You can use these indicators to judge the training progress of the model. If you find that the precision is decreasing or the model is overfitting, you can further optimize the training process by adjusting the learning rate, increasing the data set, and enhancing the data.

[0133] 7. Model evaluation and tuning:

[0134] After training, you can use the validation set to evaluate the effect of the model. If the model precision is not ideal, you can try the following methods for tuning:

[0135] Increase the diversity of the data set: Increase more images of cracks, especially in different lighting conditions, angles, and crack shapes.

[0136] Data augmentation: Perform image augmentation such as rotation, scaling, brightness adjustment, etc. to help the model better generalize.

[0137] Adjust hyperparameters: such as learning rate, batch size, training rounds, etc.

[0138] Fine-tuning: Fine-tune the model using a smaller learning rate and train for more rounds.

[0139] 8. Inference and testing

[0140] After training is complete, the trained model can be used for inference (prediction):

[0141] yolo task=detect mode=predict model=path / to / best_model.pt source=path / to / test / images

[0142] The source parameter can be the path to a test image or video file. After prediction, YOLOv8 outputs an image or video with crack labels, which can be used for further analysis.

[0143] 9. Saving and deployment

[0144] When the model training is complete and the test is passed, the model can be saved to a specified directory and deployed according to requirements. Save the model:

[0145] yolo task=detect mode=export model=path / to / best_model.pt format=onnx

[0146] The model can be exported in ONNX format for easy deployment to other platforms.

[0147] Preferably, as Figure 2 , the oblique image is preprocessed, including:

[0148] The oblique image is subjected to distortion correction processing to obtain a distortion-corrected image.

[0149] The distortion-corrected image is subjected to illumination correction processing to obtain an illumination-corrected image.

[0150] The illumination-corrected image is subjected to noise reduction processing to obtain a preprocessed oblique image.

[0151] Specifically, when the oblique image is subjected to distortion correction processing, a perspective transformation correction method can be used for correction. By calculating the homography matrix between the original image and the target image, the perspective distortion of the image can be corrected.

[0152] Preferably, the distortion-corrected image is subjected to illumination correction processing to obtain an illumination-corrected image, including:

[0153] The distortion-corrected image is partitioned using an adaptive partitioning algorithm to obtain a plurality of to-be-corrected regions.

[0154] Each to-be-corrected region is corrected separately to obtain a corrected region.

[0155] All corrected regions are merged to obtain an illumination-corrected image.

[0156] The present application can effectively reduce the probability of the occurrence of the event of enhancing the noise of the local area during the illumination correction by the way of the regional division, because the noise of the pixel of the corrected local area is easily enhanced when the illumination correction is directly performed on the distortion correction image, thereby affecting the accuracy of the subsequent recognition.

[0157] Preferably, the distortion correction image is divided by using the adaptive partition algorithm to obtain a plurality of to-be-corrected areas, comprising:

[0158] In the first step, the distortion correction image is divided into Z local areas with the same area, and the obtained local areas are stored in a set ZU.

[0159] In the second step, it is respectively judged whether each local area in ZU needs to be further divided.

[0160] In the third step, the local area needing to be further divided is stored in a set SU, and the local area not needing to be further divided is stored in a set FU.

[0161] In the fourth step, it is judged whether the number of the local areas in SU is greater than or equal to 1, if yes, the fifth step is entered, if not, the local area in FU is taken as the to-be-corrected area.

[0162] In the fifth step, ZU is reset to an empty set.

[0163] In the sixth step, each element in SU is respectively divided into Z local areas with the same area, the obtained local areas are stored in ZU, and the second step is entered.

[0164] The present application does not directly divide the distortion correction image into a plurality of to-be-corrected areas with the same area, because the size of the area of the to-be-corrected area is not easy to determine, if the area is too small, the area with low contrast is easily divided into too many to-be-corrected areas, so that the probability of the introduction of the edge information after the correction is increased, if the area is too large, a plurality of areas with large contrast difference are easily divided into the same correction area, so that the original edge information is suppressed after the illumination correction.

[0165] Therefore, the present application realizes the adaptive determination of the area of the to-be-corrected area by respectively judging whether each local area needs to be further divided, so that more edge information can be retained after the illumination correction, the introduction of the false edge information is suppressed, and the illumination correction image with higher quality is obtained.

[0166] Preferably, it is respectively judged whether each local area in ZU needs to be further divided, comprising:

[0167] denote a local region in ZU, calculate the judgment value of zu:

[0168]

[0169] Value zu denote the judgment value of zu, nzu denotes the number of pixel points in zu, gb i denote the gray value of pixel point i in zu, gbmdenotes the maximum value of the gray value of pixel points in zu; N1 denotes the number of local regions belonging to FU adjacent to zu, N2 denotes the number of local regions belonging to ZU adjacent to zu; α denotes the weight;

[0170] the judgment value of Value zu is greater than the set contrast value of the judgment value, if yes, it indicates that zu needs to be further divided, if no, it indicates that zu does not need to be further divided.

[0171] The judgment value of the application is calculated from the variation degree of the gray value of the pixel points in the local region and the number of local regions not belonging to FU around zu. When the variation degree of the gray value of the pixel points in the local region is greater and the number of local regions not belonging to FU around zu is more, the probability that zu needs to be further divided is lower. In this way, the variation degree of the gray value of the pixel points in the final obtained to-be-corrected region is smaller, the accuracy of the subsequent correction result is improved, the number of the final obtained to-be-corrected regions is effectively inhibited, and the correction efficiency is improved.

[0172] Preferably, the value of α can be 0.6.

[0173] Preferably, the set contrast value of the judgment value is 0.3.

[0174] Preferably, each to-be-corrected region is corrected respectively to obtain a correction region, including:

[0175] Each to-be-corrected region is corrected respectively using a histogram equalization algorithm to obtain a correction region.

[0176] Preferably, the preprocessed inclined plane image is spliced to obtain a spliced image, including:

[0177] The image splicing is performed using the following algorithm to obtain a spliced image:

[0178] Feature point detection is performed on two images p1 and p2 to be spliced respectively to obtain feature points in p1 and p2;

[0179] The same feature points in p1 and p2 are obtained using a matching algorithm to obtain a plurality of feature point pairs;

[0180] Eliminate the wrong feature point pairs in the plurality of feature point pairs to obtain feature point pairs for image registration;

[0181] Estimate a geometric transformation matrix between p1 and p2 based on the feature point pairs for image registration;

[0182] Transform p1 into a coordinate system in which p2 is located based on the geometric transformation matrix to obtain an image p3, and perform weighted average calculation on the gray values of the pixel points in p2 and p3 to obtain the gray value of a pixel point in a spliced image.

[0183] Preferably, the weighted average calculation on the gray values of the pixel points in p2 and p3 to obtain the gray value of a pixel point in a spliced image comprises:

[0184] For a pixel point with coordinates (x, y) in the spliced image, the gray value of the pixel point is represented as G(x, y), if there are pixel points with coordinates (x, y) in both p2 and p3, then:

[0185] G(x, y) = w1 × p2(x, y) + w2 × p3(x, y)

[0186] p2(x, y) represents the gray value of the pixel point with coordinates (x, y) in p2, p3(x, y) represents the gray value of the pixel point with coordinates (x, y) in p3, w1 and w2 represent the first weighting coefficient and the second weighting coefficient, respectively.

[0187] If there is a pixel point with coordinates (x, y) in p2, and there is no pixel point with coordinates (x, y) in p3, then G(x, y) = p2(x, y);

[0188] If there is a pixel point with coordinates (x, y) in p3, and there is no pixel point with coordinates (x, y) in p2, then G(x, y) = p3(x, y).

[0189] After converting p1 into the coordinate system in which p2 is located, some pixel points in the obtained image p3 will have the same coordinates as the pixel points of p2, so it is necessary to determine the gray values of these pixel points. The present application can make the transition of the boundary of the spliced image more natural and retain more edge information by performing weighted summation on the gray values of the pixel points corresponding to the same coordinates in p2 and p3 through the first weighting coefficient and the second weighting coefficient.

[0190] Preferably, the calculation formula of w1 is:

[0191]

[0192] w2 = 1 - w1

[0193] d2 represents the distance between the pixel point with coordinates (x, y) and the edge L3, d3 represents the distance between the pixel point with coordinates (x, y) and the edge L2, p23 represents the set of coordinates of the pixel points in the overlapping area between p2 and p3; pl2 represents the area composed of the pixel points in p2 with coordinates not belonging to p23; L2 is the boundary line between p23 and pl2; pl3 represents the area composed of the pixel points in p3 with coordinates not belonging to p23; L3 is the boundary line between p23 and pl3, λ represents a reference ratio, neg2 represents the total number of edge pixel points contained in the area composed of the pixel points corresponding to the coordinates in the set p23 in p2; and neg3 represents the total number of edge pixel points contained in the area composed of the pixel points corresponding to the coordinates in the set p23 in p3.

[0194] In the calculation of w1, not only the distance between the pixel point with coordinates (x, y) and the edge of the overlapping area is considered, but also the total number of edge pixel points contained in the area corresponding to the coordinates of the pixel points in the overlapping area in p2 is considered, so that the influence of p2 (x, y) on the result of weighting is greater when the total number of edge pixel points contained in the area corresponding to the coordinates of the pixel points in the overlapping area in p2 is greater and the value of d2 is greater, the image edge information of the image after splicing can be retained while the edge transition is kept natural, so that the accuracy of subsequent crack identification is improved.

[0195] Preferably, the value of λ is 0.7.

[0196] Preferably, the matching algorithm includes any one of FLANN and KNN.

[0197] Preferably, the false feature point pairs in the plurality of feature point pairs are removed to obtain the feature point pairs for image registration, including:

[0198] The false feature point pairs in the plurality of feature point pairs are removed using the RANSAC algorithm to obtain the feature point pairs for image registration.

[0199] The preferred embodiments disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details, nor limit the application to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of the present application. The embodiments are selected and described in detail in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited only by the claims and their full scope and equivalents.< / height> < / width>

Claims

1. A dam crack detection system based on unmanned aerial vehicle technology, characterized in that, The unmanned aerial vehicle and the image processing terminal are included; The unmanned aerial vehicle is used for photographing the slope of the dam to obtain a plurality of slope images; The image processing terminal is used for identifying the slope images to obtain a detection result, including: Each of the slope images is preprocessed to obtain a preprocessed slope image; The preprocessed slope images are spliced to obtain a spliced image; A detection algorithm based on deep learning is used to identify the spliced image to obtain a detection result; Wherein, photographing the slope of the dam includes: Based on the height of the slope of the dam, a plurality of photographing heights are obtained; The unmanned aerial vehicle moves horizontally at each photographing height, and during the movement, the distance between the unmanned aerial vehicle and the slope of the dam in the horizontal direction is kept the same, and the slope of the dam is photographed once every fixed movement distance to obtain a slope image; The slope image is preprocessed, including: The slope image is subjected to distortion correction processing to obtain a distortion corrected image; The distortion corrected image is subjected to illumination correction processing to obtain an illumination corrected image; The illumination corrected image is subjected to noise reduction processing to obtain a preprocessed slope image; The illumination corrected image is obtained by performing illumination correction processing on the distortion corrected image, including: An adaptive partitioning algorithm is used to partition the distortion corrected image to obtain a plurality of to-be-corrected regions; Each of the to-be-corrected regions is corrected to obtain a corrected region; All the corrected regions are merged to obtain an illumination corrected image; An adaptive partitioning algorithm is used to partition the distortion corrected image to obtain a plurality of to-be-corrected regions, including: Firstly, the distortion corrected image is divided into Z local regions with the same area, and the obtained local regions are stored in a set ZU; Secondly, it is judged whether each local region in ZU needs to be further divided; Thirdly, the local regions that need to be further divided are stored in a set SU, and the local regions that do not need to be further divided are stored in a set FU; Fourthly, it is judged whether the number of local regions in SU is greater than or equal to 1, if yes, the fifth step is entered, if not, the local regions in FU are taken as to-be-corrected regions; Fifthly, ZU is reset to an empty set; Sixthly, each element in SU is divided into Z local regions with the same area, and the obtained local regions are stored in a set ZU, and the second step is entered.

2. The dam crack detection system based on unmanned aerial vehicle technology according to claim 1, characterized in that, The image processing terminal is a computer, which can establish a connection with the unmanned aerial vehicle through a WiFi connection or a data line connection, and after the connection is established, the computer can access the photographed slope images stored in the storage device of the unmanned aerial vehicle. 3.The dam crack detection system based on UAV technology of claim 1, wherein, Based on the height of the slope of the dam, a plurality of photographing heights are obtained, including: H represents the height of the slope of the dam; D represents the distance between the unmanned aerial vehicle and the slope of the dam in the horizontal direction during photographing; N represents the number of photographing heights, which can enable the unmanned aerial vehicle to photograph the slope of the dam based on D, and the slope images obtained at adjacent photographing heights have overlapping regions in the vertical direction; then the first photographing height is the n-th photographing height is 4. The dam crack detection system based on unmanned aerial vehicle technology according to claim 3, characterized in that, The relationship between the area of the overlapping region and the area of the slope image is: Sa represents the area of the overlapping region in the slope image, Sb represents the area of the slope image, and h represents a set proportion. 5.The dam crack detection system based on UAV technology of claim 4, wherein, The value of h is 0.

5.

6. The dam crack detection system based on UAV technology according to claim 1, wherein, The preprocessed inclined plane images are spliced to obtain a spliced image, comprising: The image splicing is performed using the following algorithm to obtain a spliced image: Feature point detection is performed on two images p1 and p2 to be spliced to obtain feature points in p1 and p2; The same feature points in p1 and p2 are obtained using a matching algorithm to obtain a plurality of feature point pairs; The incorrect feature point pairs in the plurality of feature point pairs are removed to obtain feature point pairs for image registration; A geometric transformation matrix between p1 and p2 is estimated based on the feature point pairs for image registration; p1 is transformed into the coordinate system of p2 based on the geometric transformation matrix to obtain an image p3, and the gray values of the pixel points in p2 and p3 are calculated by weighted average in the coordinate system of p2 to obtain the gray values of the pixel points in the spliced image. 7.The dam crack detection system based on UAV technology of claim 6, wherein, The matching algorithm includes any one of FLANN and KNN. 8.The dam crack detection system based on UAV technology of claim 6, wherein, The incorrect feature point pairs in the plurality of feature point pairs are removed to obtain feature point pairs for image registration, comprising: The incorrect feature point pairs in the plurality of feature point pairs are removed using a RANSAC algorithm to obtain feature point pairs for image registration.

Images