A deformation detection method and system for river flood control wall

Abstract

The present invention discloses a deformation detection method and system for a river flood control wall, which determines a corresponding warning data set according to a vertical settlement parameter, a lateral displacement parameter, and a corresponding warning threshold, and is compatible with multiple interactions of the vertical settlement parameter, the lateral displacement parameter, and the corresponding warning threshold, thereby ensuring the accuracy of the warning data set, and thus ensuring the accuracy of the warning image. Furthermore, multiple judgments are made on non-similar feature sets and warning data sets, and the recognition results of the non-similar feature sets and the deformation results of the warning data sets are output. The deformation results of the river flood control wall are determined according to the recognition results of the non-similar feature sets, the deformation results of the warning data sets, and the result matching table, realizing multi-dimensional control of the recognition results of the non-similar feature sets, the deformation results of the warning data sets, and the result matching table, ensuring multi-level control of the deformation results of the river flood control wall, and thus ensuring the accuracy of the deformation results of the river flood control wall.

Description

Technical Field

[0001] The present invention relates to the technical field of river flood control walls, and in particular to a deformation detection method and system for river flood control walls. Background Art

[0002] With the development of science and technology, river flood control walls, as important flood control facilities, play an important role in ensuring urban flood control safety. Deformation monitoring of flood control walls is the top priority of urban river management. Most urban river flood control walls adopt gravity flood control wall structures, and deformation damage is one of the main hazards in the operation of such flood control walls.

[0003] At present, the detection of deformation and damage of this type of flood control wall mainly relies on manual inspections and expert experience, or online monitoring of deformation and damage with the help of GNSS monitoring stations and other equipment, combined with manual on-site verification methods, and the accuracy of the early warning images cannot be guaranteed, which leads to low accuracy of the deformation results of river flood control walls. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art. The present invention provides a deformation detection method and system for a river flood control wall.

[0005] The embodiment of the present invention provides a deformation detection method for a river flood control wall, which is applied to the deformation detection scenario of a river flood control wall;

[0006] The deformation detection method of the river flood control wall comprises: collecting a real-time image of the river flood control wall; determining a corresponding first feature set according to the real-time image of the river flood control wall and a recognition model; determining a vertical settlement parameter and a lateral displacement parameter based on the deformation detection of the river flood control wall by a GNSS displacement monitoring station; determining a corresponding early warning data set according to the vertical settlement parameter, the lateral displacement parameter and a corresponding early warning threshold, and triggering the collection of a corresponding early warning image according to the early warning data set; determining a corresponding second feature set based on the early warning image and the recognition model, and determining a non-similar feature set according to the comparison between the second feature set and the first feature set; performing multiple judgments on the non-similar feature set and the early warning data set, and outputting the recognition result of the non-similar feature set and the deformation result of the early warning data set, and determining the deformation result of the river flood control wall according to the recognition result of the non-similar feature set, the deformation result of the early warning data set and a result matching table.

[0007] In addition, an embodiment of the present invention further provides a deformation detection system for a river flood control wall, the deformation detection system for a river flood control wall comprising: an acquisition module for acquiring a real-time image of the river flood control wall; a first feature module for determining a corresponding first feature set based on the real-time image of the river flood control wall and a recognition model; a deformation detection module for determining a vertical settlement parameter and a lateral displacement parameter based on deformation detection of the river flood control wall by a GNSS displacement monitoring station; an early warning module for determining a corresponding early warning data set based on the vertical settlement parameter, the lateral displacement parameter, and a corresponding early warning threshold, and triggering the acquisition of a corresponding early warning image based on the early warning data set; a comparison module for determining a corresponding second feature set based on the early warning image and the recognition model, and determining a non-similar feature set based on the comparison between the second feature set and the first feature set; a result matching module for performing multiple judgments on the non-similar feature set and the early warning data set, and outputting the recognition result of the non-similar feature set and the deformation result of the early warning data set, and determining the deformation result of the river flood control wall based on the recognition result of the non-similar feature set, the deformation result of the early warning data set, and the result matching table.

[0008] In an embodiment of the present invention, through the method in the embodiment of the present invention, a real-time image of a river flood control wall is collected; a corresponding first feature set is determined based on the real-time image of the river flood control wall and a recognition model; vertical settlement parameters and lateral displacement parameters are determined based on deformation detection of the river flood control wall by a GNSS displacement monitoring station; a corresponding warning data set is determined based on the vertical settlement parameters, the lateral displacement parameters, and the corresponding warning thresholds, and the collection of the corresponding warning image is triggered based on the warning data set, which is compatible with multiple interactions of the vertical settlement parameters, the lateral displacement parameters, and the corresponding warning thresholds, thereby ensuring the accuracy of the warning data set and thus ensuring the accuracy of the warning image.

[0009] Furthermore, a corresponding second feature set is determined based on the warning image and the recognition model, and a non-similar feature set is determined based on the comparison between the second feature set and the first feature set; multiple judgments are made on the non-similar feature set and the warning data set, and the recognition results of the non-similar feature set and the deformation results of the warning data set are output; the deformation results of the river flood control wall are determined based on the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table, which are compatible with the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table, and realize multi-dimensional control of the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table, ensuring multi-level control of the deformation results of the river flood control wall, and then ensuring the accuracy of the deformation results of the river flood control wall, realizing all-weather, timely, efficient and intelligent deformation detection of the river flood control wall, thereby improving the river risk prevention and control capabilities and providing strong guarantees for urban flood control safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0011] Figure 1 is a flow chart of a deformation detection method for a river flood control wall in an embodiment of the present invention;

[0012] Figure 2 Schematic diagram of the structural composition of a deformation detection system for a river flood control wall in an embodiment of the present invention;

[0013] Figure 3 The figure is a hardware diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0014] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0015] See also Figures 1 to 3 , a deformation detection method for a river flood control wall, applied to a deformation detection scenario of a river flood control wall; the deformation detection method for a river flood control wall includes:

[0016] Step S11: collecting real-time images of the river flood control wall;

[0017] Step S12: determining a corresponding first feature set according to the real-time image of the river flood control wall and the recognition model;

[0018] Step S13: determining vertical settlement parameters and lateral displacement parameters based on deformation detection of the river flood control wall by the GNSS displacement monitoring station;

[0019] Step S14: determining a corresponding warning data set according to the vertical settlement parameter, the lateral displacement parameter, and the corresponding warning threshold, and triggering the collection of a corresponding warning image according to the warning data set;

[0020] Step S15: determining a corresponding second feature set based on the warning image and the recognition model, and determining a non-similar feature set based on a comparison between the second feature set and the first feature set;

[0021] Step S16: Perform multiple judgments on the non-similar feature set and the warning data set, and output the recognition results of the non-similar feature set and the deformation results of the warning data set. Determine the deformation results of the river flood control wall based on the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table.

[0022] In step S11, a real-time image of the river flood control wall is collected;

[0023] In the specific implementation process of the present invention, the specific steps may be:

[0024] S111: Collect the location of the river flood control wall;

[0025] S112: determining a corresponding flight detection area based on the location of the river flood control wall;

[0026] S113: Determine the flight path of the UAV according to the environmental parameters of the flight detection area and the river flood control wall;

[0027] S114: triggering the dynamic flight of the drone based on the flight path of the drone, and the drone dynamically photographing the river flood control wall during the dynamic flight;

[0028] S115: Collecting a real-time image of the river flood control wall based on the dynamic shooting of the river flood control wall by the drone.

[0029] In an embodiment of the present application, the location of the river flood control wall is collected and introduced, and the location of the river flood control wall is controlled, so as to determine the corresponding flight detection area based on the location of the river flood control wall, so as to facilitate subsequent control of the flight detection area.

[0030] At this time, the exact location of the river flood control wall is determined to provide basic data for subsequent flight detection area planning and drone flight path planning. Use a GPS locator or similar equipment to locate the river flood control wall site and obtain accurate latitude and longitude information.

[0031] According to the location and length of the flood control wall, determine a suitable flight detection area to ensure that the drone can fully cover the flood control wall for shooting. According to the collected flood control wall location information, determine its length, width and shape. On the basis of the flood control wall, expand a certain range (such as 50 meters or 100 meters) outward to form a flight detection area that includes the flood control wall and has a certain safety margin. This area should be large enough to accommodate the flight and shooting needs of the drone, but small enough to reduce unnecessary flight time and energy consumption. When determining the flight detection area, it is also necessary to consider factors in the surrounding environment, such as obstacles such as buildings, trees, and wires, as well as meteorological conditions such as wind direction and wind speed. These factors may affect the flight safety and shooting effect of the drone, so they need to be fully considered during planning.

[0032] Furthermore, the flight path of the UAV is determined according to the environmental parameters of the flight detection area and the river flood control wall. The environmental parameters of the flight detection area and the river flood control wall are introduced, and the environmental parameters of the flight detection area and the river flood control wall are considered as a whole. This realizes multi-dimensional control of the environmental parameters of the flight detection area and the river flood control wall, ensuring the accuracy of the UAV's flight path.

[0033] At this time, plan the flight path of the drone to ensure that the drone can fully cover the river flood control wall in the flight detection area and capture high-quality image data. First, understand the shape, size and boundary conditions of the flight detection area. This helps to determine the basic parameters of the drone, such as the take-off point, flight altitude, and flight speed. Analyze the structural characteristics of the river flood control wall, possible safety hazards (such as cracks, deformation, etc.), and the image resolution and shooting angle required to shoot these hazards. These factors will affect the flight path planning of the drone to ensure that key information can be captured. Comprehensively consider the environmental parameters in the flight detection area, such as terrain undulations, obstacle distribution (such as trees, buildings), wind direction and speed, electromagnetic interference, etc. These factors may have a direct impact on the flight safety and shooting effect of the drone, so they need to be fully considered when planning the flight path.

[0034] Select the appropriate flight mode according to actual needs and environmental conditions. For example, for complex flight environments, you can choose the automatic obstacle avoidance mode; for areas that require high-precision shooting, you can choose the fixed-point hovering mode; for long-distance flights, you can consider the energy-saving flight mode, etc. Based on the above analysis, use the UAV flight planning software or algorithm, combined with the performance parameters of the UAV (such as maximum flight altitude, speed, endurance, etc.), to plan the optimal flight path. This path should be able to fully cover the flight detection area while ensuring the safe flight and efficient shooting of the UAV. After the planning is completed, perform a flight simulation to check the feasibility and safety of the path. If potential problems are found (such as collision risks, unstable flight, etc.), the flight path needs to be adjusted and optimized.

[0035] Specifically, suppose we need to inspect a section of flood control wall located along a river in the city. First, we determined the flight inspection area, which extends about 2 kilometers along the flood control wall and is 100 meters wide. Considering the structural characteristics and safety hazards of the flood control wall, we decided to take high-resolution photos at key locations (such as cracks and deformations). When planning the flight path, we noticed that there were several high-rise buildings and several tall trees in the flight inspection area. In order to avoid collisions with these obstacles, we selected the automatic obstacle avoidance mode and used these obstacles as no-fly zones when planning the path. At the same time, considering the influence of wind direction and speed, we adjusted the flight direction so that the drone could fly downwind to reduce flight resistance. After determining the flight altitude, speed and shooting angle, we used the drone flight planning software to plan the optimal flight path. The path starts from the starting point of the flood control wall, flies along its extension direction, and hovers at key locations for shooting. To ensure safety, we performed a flight simulation before takeoff and adjusted the path to avoid potential risks. In the end, the drone completed the flight mission safely and efficiently according to the planned path, and captured high-quality image data of the river flood control wall. These data provide strong support for subsequent safety monitoring and analysis.

[0036] In another embodiment of the present application, the matching table method is to determine the most suitable flight path by matching the environmental parameters of the flight detection area and the river flood control wall with the flight performance parameters of the drone. The following is a simple matching table example:

[0037] Flight detection area characteristics Environmental parameters UAV performance parameters Flight path suggestions Straight river channel No tall obstacles High maximum flight speed Fly in a straight line along the river, maintaining a constant speed Curved River There are a few trees Strong obstacle avoidance capability Fly along the bend of the river and avoid trees Narrow river Strong electromagnetic interference Strong anti-interference ability Keep low altitude flight to reduce electromagnetic interference Wide river Large changes in wind direction and speed Strong wind resistance When flying against the wind or in a crosswind, adjust the flight altitude and speed appropriately.

[0038] In this example, we give corresponding flight path suggestions based on the shape of the flight detection area, environmental parameters, and the performance parameters of the drone. For example, in a straight river without tall obstacles, it is recommended that the drone fly in a straight line along the river and maintain a constant speed; in a curved river with a few trees, it is recommended that the drone fly along the curved river and pay attention to avoid the trees.

[0039] Therefore, the dynamic flight of the UAV is triggered based on the flight path of the UAV, and the UAV dynamically photographs the river flood control wall during the dynamic flight; real-time images of the river flood control wall are collected based on the dynamic photography of the river flood control wall by the UAV.

[0040] In this embodiment, a drone is used to fly at a speed of 1m / s. The drone flies along the flight path. At the same time, the drone uses the onboard camera to shoot the area 10 meters around the GNSS installation point of the river channel to obtain 10 image data sets from the front and top perspectives. At this time, the real-time image of the river flood control wall is used as the image data set.

[0041] At this point, start the drone and perform dynamic flight according to the pre-planned flight path, while ensuring that the drone can take continuous and stable dynamic shots of the river flood control wall during the flight. Upload the flight path planned in step S113 to the drone's control system. This is usually done through the drone ground station software or mobile application. Before starting the drone, make sure that the drone is in good condition, the battery is fully charged, and the camera and sensors are working properly. At the same time, check the safety of the take-off and landing points to ensure that there are no obstacles that affect the take-off and landing of the drone.

[0042] Start the drone according to the drone's operating manual and put it in the take-off position. Issue a take-off command through the ground station or remote control, and the drone will start flying according to the preset flight path. During the flight, the drone's camera will automatically or manually adjust to the appropriate angle and focal length to continuously shoot the river flood control wall. Shooting parameters (such as exposure time, shutter speed, ISO, etc.) may need to be adjusted in real time according to lighting conditions to ensure image quality. Throughout the flight, the flight status of the drone, including flight altitude, speed, position, and battery power, is continuously monitored through the ground station or remote control. If any abnormality is found, appropriate measures should be taken immediately, such as adjusting the flight path, reducing the flight speed, or emergency landing.

[0043] In step S12, a corresponding first feature set is determined according to the real-time image of the river flood control wall and the recognition model;

[0044] In the specific implementation process of the present invention, the specific steps may be:

[0045] S121: freeze-frame real-time image of river flood control wall;

[0046] S122: determining a grayscale image based on the real-time image of the river flood control wall and a perceptual weighted grayscale algorithm;

[0047] S123: triggering corresponding noise reduction according to the grayscale image and the Gaussian blur algorithm, and outputting the image after noise reduction processing;

[0048] S124: Associating the image after noise reduction processing with the recognition model;

[0049] S125: Determine a corresponding first feature set according to the image after denoising and the recognition model.

[0050] In an embodiment of the present application, the real-time image of the river flood control wall is frozen, the real-time image of the river flood control wall is introduced, and the real-time image of the river flood control wall is managed and controlled. At the same time, the grayscale image is determined based on the real-time image of the river flood control wall and the perceptual weighted grayscale algorithm, and the real-time image of the river flood control wall and the perceptual weighted grayscale algorithm are compatible. The real-time image of the river flood control wall and the perceptual weighted grayscale algorithm are controlled as a whole to ensure the accuracy of the grayscale image.

[0051] At this point, the most representative image, i.e., the freeze-frame image, is selected from the continuous real-time images. This usually involves judging the image quality and capturing the state of the river flood control wall. In order to obtain high-quality freeze-frame images, some technical means may need to be taken, such as adjusting the camera's focal length, exposure time, white balance and other parameters to ensure that the image is clear and the color is accurate.

[0052] In practice, this step may be completed automatically by a camera on a drone, usually relying on an image recognition algorithm that can determine when the captured image best meets preset quality standards or best reflects the status of the river flood control wall.

[0053] Convert the frozen color image to a grayscale image. Grayscale images have a lower data volume than color images, so they can reduce the amount of calculation and increase the processing speed when performing image processing. In addition, grayscale images can also show the contour and texture information of the image more clearly, which is of great significance for subsequent image analysis and feature extraction. The perceptual weighted grayscale algorithm is a commonly used grayscale method. It assigns different weights to different color channels according to the different sensitivities of the human visual system to different colors. In a color image, the color of each pixel is composed of three color channels: red, green, and blue. The perceptual weighted grayscale algorithm will synthesize them into a grayscale value according to certain weights based on the values ​​of these three color channels. In actual operation, this step is usually implemented through image processing software or algorithm library. First, the frozen color image is input into the grayscale algorithm; then, the algorithm will perform weighted summation of the three color channels of red, green, and blue according to the preset weights to obtain the grayscale value of each pixel; finally, these grayscale values ​​are combined to form a grayscale image.

[0054] Specifically, suppose that in a monitoring mission of a river flood control wall, the drone continuously captures real-time images of the river flood control wall according to the preset flight path. When flying to a certain location, the camera captures a clear image in which the structure of the river flood control wall is clearly visible without interference from factors such as clouds, shadows or reflections. At this time, the image recognition algorithm determines that the image quality is high and meets the freeze standard, so it automatically triggers the freeze operation and saves the image as input for subsequent processing. After freezing a clear color image, we input it into the perceptual weighted grayscale algorithm. The algorithm performs weighted summation of the red, green and blue color channels according to the preset weights (such as red 0.299, green 0.587, blue 0.114) to obtain the grayscale value of each pixel. Finally, these grayscale values ​​are combined to form a clear grayscale image. This image retains the contour and texture information of the original color image while reducing the amount of data, which facilitates subsequent image analysis and feature extraction.

[0055] Furthermore, according to the grayscale image and Gaussian blur algorithm, the corresponding denoising is triggered, and the denoised image is output; the denoised image and the recognition model are associated; the corresponding first feature set is determined according to the denoised image and the recognition model, which is compatible with the overall consideration of the denoised image and the recognition model, realizes the accurate recognition of the denoised image and the recognition model, and ensures the accuracy of the first feature set.

[0056] At this time, the grayscale image is subjected to noise reduction. Noise reduction is an important part of image processing. It can remove noise from the image, improve image quality, and provide more accurate data for subsequent feature extraction and recognition. Gaussian blur algorithm is a commonly used noise reduction method. It reduces the impact of noise on the image by weighted averaging each pixel in the image. The weight of the weighted average is determined by a Gaussian function, which assigns weights according to the distance between the pixel and the center pixel. The closer the distance, the greater the weight of the pixel, and the farther the distance, the smaller the weight of the pixel. In actual operation, this step is usually implemented by image processing software or algorithm library. First, the grayscale image is input into the Gaussian blur algorithm; then, the image is blurred according to the preset blur degree (i.e., the size and standard deviation of the Gaussian kernel) to remove noise; finally, the image after noise reduction is output.

[0057] Associate the denoised image with a pre-trained recognition model. A recognition model is a machine learning model that extracts useful features from an image and classifies or identifies the image based on those features. In practice, this step usually involves feeding the denoised image into the recognition model and waiting for the model to output the recognition result. The recognition model may have been trained to recognize specific features in the river flood control wall image, such as cracks, spalling, deformation, etc.

[0058] According to the image after denoising and the output result of the recognition model, the corresponding first feature set is determined. The first feature set is a set of key features extracted from the image, which can reflect the status and safety of the river flood control wall. In actual operation, this step usually involves parsing and extracting the output results of the recognition model to obtain the values ​​of the key features. These features may include the length, width, depth of the crack, the area and position of the peeling, the degree of deformation, etc. Then, these features are combined into a feature set, namely the first feature set.

[0059] Specifically, suppose we have obtained a grayscale image of a river flood control wall, but there is some noise in the image due to camera shaking or environmental factors. In order to remove this noise, we input the image into the Gaussian blur algorithm and set the Gaussian kernel size to 5x5 and the standard deviation to 1.0. The algorithm performs a weighted average on each pixel in the image, removes the noise, and outputs a denoised image. The processed image is clearer and the noise is significantly reduced, providing more accurate data for subsequent feature extraction and recognition.

[0060] Assume that we have trained a recognition model that can identify crack features in river flood control wall images. Now, we input the denoised image into the model and wait for the model to output the recognition result. The model extracts and classifies the image and finally outputs detailed information such as the location, size and shape of the crack. This information is important for assessing the safety of river flood control walls and formulating maintenance plans. The crack features are identified from the denoised image, and detailed information such as the location, size and shape of the crack is obtained. Now, we combine this information into a feature set, namely the first feature set. This set includes key features such as the length (such as 5 meters), width (such as 0.2 meters), and depth (such as 0.1 meters) of the crack. These features are of great value for subsequent river flood control wall status assessment and hidden danger identification.

[0061] In this embodiment, the real-time image of the river flood control wall is used as the basic data of the river. The image is grayed using the perceptual weighted graying algorithm. The three channels (red, green, and blue) of the color image are grayed. The calculation formula is:

[0062] ,

[0063] I gray is the pixel value in the grayscale image.

[0064] R is the pixel value of the red channel.

[0065] G is the pixel value of the green channel.

[0066] B is the pixel value of the blue channel.

[0067] (0.289, 0.687, 0.304) The weights are obtained after multiple tests.

[0068] At the same time, the Gaussian blur algorithm is used to reduce the noise of the grayscale image. The Gaussian blur algorithm makes the pixel value more stable by smoothing the image, thereby improving the accuracy of feature point detection.

[0069] Determine a Gaussian kernel of size 7x7, apply the Gaussian kernel to each pixel on the grayscale image, and implement Gaussian blur convolution through Gaussian blur algorithm formula operation. The Gaussian blur algorithm formula is as follows:

[0070] ,

[0071] I gray (x,y) is the pixel value of the original grayscale image at position (x,y).

[0072] H(i,j) is the value of the Gaussian kernel at position (i,j).

[0073] k is the Gaussian kernel radius.

[0074] Furthermore, the number of pixels of the image after noise reduction is doubled and upsampled to form the first group of sampling images, and then the first group of images is downsampled using the Gaussian blur convolution algorithm using the determined 7x7 Gaussian kernel to obtain the second group of sampling images. The subsequent downsampling operation is the same as the second group until only one pixel is left in the last group. Based on the above sampling data, a Gaussian pyramid with 5 layers is constructed in each group to obtain the feature information of the target object. The formula for constructing the Gaussian pyramid is:

[0075] ,

[0076] O: The number of Gaussian pyramid groups.

[0077] n: The number of layers extracted in the final Difference of Gaussian pyramid.

[0078] o: The index number of the group, o∈[0,1....,n-1].

[0079] r: The index number of the group, r∈[0,1....,n+2].

[0080] σ(o, r): corresponds to the Gaussian blur coefficient of the image.

[0081] Subtract each group of adjacent layers of the Gaussian pyramid to obtain the scale space composed of the Gaussian difference pyramid. Start searching from the second layer of the Gaussian difference pyramid. Compare the middle detection point and its 8 adjacent points of the same scale with the 9×2 points corresponding to the upper and lower adjacent scales, a total of 26 points, and find the maximum value as the potential key point. Perform a ternary second-order Taylor expansion x0 (x0, y0, σ0) on the found potential key point, find the extreme value and differentiate it to get its rate of change. When the rate of change is 0, it means that the true extreme point has been found, and the extreme point is recorded as the feature information point. All feature information points together constitute the feature information of the target object.

[0082] Get the gradient direction and gradient amplitude of all feature information points in the feature information of the target object. The specific method is: divide 0~360° into 36 columns, each column is 10°. Find the position corresponding to the key point in the Gaussian pyramid, use it as the center of the Gaussian difference pyramid, and draw a circle with a radius of 1.88 times the scale σ. Count the gradient direction and gradient amplitude of all pixels in the circle, and use 1.88σ for Gaussian filtering. By Gaussian weighting the gradient amplitude of each point, the gradient amplitude near the feature point has a larger weight, which can partially compensate for the problem of unstable feature points caused by the lack of affine invariance. Finally, the direction with the highest value is counted as the main direction, and the direction with a value greater than 90% of the main direction is retained as the auxiliary direction.

[0083] Generate a feature point descriptor vector. Similar to the method of determining the main direction, divide the neighborhood near the feature point into d*d sub-regions, each sub-region is divided into 8 directions (each 45°), and the size of each sub-region is mσ*mσ pixels, where σ is the scale value of the feature point. Rotate the image so that the main direction of the point coincides with the x-axis of the plane rectangular coordinate system. Calculate the gradient amplitude and amplitude of the pixels in the neighborhood after rotation, and then use σ=d / 2 for Gaussian weighting to obtain the descriptor of the point. Let d=4, and the descriptor becomes a 4*4*8=128-dimensional vector.

[0084] The feature point descriptor vector is used to construct a description histogram. Since each coordinate point after rotation cannot completely coincide with the feature point, interpolation sampling is required and proportionally distributed to the two directions closest to it. Imagine the generated 128-dimensional vector as a set of cubes. Theoretically, the two directions on both sides of the direction cannot be exactly in the above 8 directions. Use coordinate translation to translate the center point to the vertex of the cube and calculate the contribution of the point in the cube to the 8 vertices of the cube. Finally, in order to remove the influence of illumination changes, normalization is performed.

[0085] In step S13, vertical settlement parameters and lateral displacement parameters are determined based on deformation detection of the river flood control wall by the GNSS displacement monitoring station;

[0086] In the specific implementation process of the present invention, the specific steps may be:

[0087] S131: Fixed-frame GNSS displacement monitoring station;

[0088] S132: GNSS displacement monitoring station detects deformation of river flood control walls;

[0089] S133: Real-time monitoring of deformation detection of river flood control walls;

[0090] S134: In the deformation detection of the river flood control wall, the detection object of the GNSS displacement monitoring station is collected;

[0091] S135: Determine vertical settlement parameters and lateral displacement parameters based on real-time detection of the detection object and the river flood control wall.

[0092] In an embodiment of the present application, a GNSS displacement monitoring station is fixed, the GNSS displacement monitoring station is controlled, and the GNSS displacement monitoring station performs deformation detection on the river flood control wall; the deformation detection of the river flood control wall is monitored in real time to ensure the deformation detection of the river flood control wall.

[0093] At this point, the location of the GNSS displacement monitoring station is selected and set to ensure that it can effectively monitor the deformation of the river flood control wall. This step involves several key elements, including the selection of monitoring sites, the installation of equipment, the determination of coordinate systems, and the connection of data transmission systems.

[0094] According to the characteristics of the monitored object (such as river flood control wall) and the monitoring requirements, select a suitable location for setting up a GNSS displacement monitoring station. Usually, these locations are selected near the surface or structure to be monitored, especially those areas that are susceptible to deformation. At the selected location, install the GNSS receiver and antenna according to the guidance of the equipment manufacturer or monitoring requirements. Ensure that the equipment is installed firmly and oriented correctly to avoid obstructions that affect the quality of the received signal. Before installing the equipment, it is necessary to determine the coordinate system of the monitoring site to ensure the accuracy and comparability of the monitoring data. Internationally accepted coordinate systems are usually used. Connect the GNSS receiver to the data acquisition system to ensure that the monitoring data can be collected in real time and transmitted to the central server or monitoring center for further processing and analysis.

[0095] Deformation detection of river flood control walls. This step involves data collection, processing and analysis to obtain deformation parameters (such as vertical settlement and lateral displacement). GNSS displacement monitoring stations receive navigation signals from multiple satellites and use differential positioning technology to achieve high-precision positioning of monitoring points. Data collection is continuous to ensure that any slight deformation can be captured. The collected data will be transmitted to the data center or monitoring center in real time for further processing and analysis. The processing process may include data filtering, error correction, etc. to improve the accuracy and reliability of the data. By analyzing the processed data, the deformation parameters of the river flood control wall can be obtained, including vertical settlement amount and rate, lateral displacement amount and direction, etc. These parameters are crucial for evaluating the stability and safety of the river flood control wall.

[0096] Therefore, in the deformation detection of the river flood control wall, the detection objects of the GNSS displacement monitoring station are collected; the vertical settlement parameters and lateral displacement parameters are determined according to the real-time detection of the detection objects and the river flood control wall, thus realizing the real-time detection of the detection objects and the river flood control wall, and ensuring the precise control of the vertical settlement parameters and lateral displacement parameters.

[0097] At this time, the deformation of the river flood control wall is monitored in real time. This usually involves an integrated monitoring system that can receive data from the GNSS displacement monitoring station in real time, perform preliminary processing, and display the deformation. The purpose of real-time monitoring is to detect abnormal deformation in a timely manner and provide data support for timely response measures.

[0098] The monitoring system receives data from GNSS displacement monitoring stations in real time, including satellite signals, location coordinates, etc. These data will undergo preliminary processing, such as denoising and filtering, to improve the accuracy and reliability of the data. The processed data will be used to calculate the deformation of the river flood control wall, including vertical settlement and lateral displacement. These calculations are usually based on differential positioning technology, and the deformation parameters are obtained by comparing the position coordinates at different time points. The calculated deformation will be displayed graphically on the monitoring interface, so that the monitoring personnel can intuitively understand the deformation trend of the river flood control wall. At the same time, if the deformation exceeds the preset threshold, the monitoring system will automatically trigger an alarm to notify relevant personnel to take countermeasures.

[0099] During the deformation detection process, the detection objects of the GNSS displacement monitoring station are identified and collected. The detection objects here usually refer to the key parts or overall structures of the river flood control wall, which are easily affected by deformation and are therefore the focus of monitoring. According to the structural characteristics and monitoring requirements of the river flood control wall, the key parts or overall structures that need to be monitored are determined. These parts may include embankments, dam bodies, slope protection, etc. GNSS displacement monitoring stations are installed on the determined detection objects, and relevant data are collected. These data include satellite signals, location coordinates, etc., which are used for subsequent deformation calculation and analysis.

[0100] The vertical settlement parameters and lateral displacement parameters of the river flood control wall are determined based on the real-time detection data. These parameters are an important basis for evaluating the stability and safety of the river flood control wall. The real-time detection data is further processed and analyzed, including data filtering, error correction, etc., to improve the accuracy and reliability of the data. Based on the processed data, the vertical settlement parameters (such as settlement amount, settlement rate) and lateral displacement parameters (such as displacement amount, displacement direction) of the river flood control wall are calculated. These parameters can be obtained by comparing the position coordinates at different time points. Based on the calculated deformation parameters, the stability and safety of the river flood control wall are evaluated. If the deformation parameters exceed the preset threshold, the early warning mechanism will be triggered to notify relevant personnel to take countermeasures.

[0101] In step S14, a corresponding warning data set is determined according to the vertical settlement parameter, the lateral displacement parameter, and the corresponding warning threshold, and the collection of the corresponding warning image is triggered according to the warning data set;

[0102] In the specific implementation process of the present invention, the specific steps may be:

[0103] S141: Fixed vertical settlement parameters and lateral displacement parameters;

[0104] S142: Associating vertical settlement parameters, lateral displacement parameters, and corresponding warning thresholds;

[0105] S143: Compare the vertical settlement parameter and the lateral displacement parameter with the corresponding warning threshold in real time;

[0106] S144: if the vertical settlement parameter and the lateral displacement parameter are greater than the corresponding warning threshold, outputting the vertical settlement parameter and the lateral displacement parameter in the warning state;

[0107] S145: determining a corresponding warning data set based on the vertical settlement parameter and the lateral displacement parameter in the warning state;

[0108] S146: triggering the drone to detect again according to the warning data set, and outputting the corresponding warning image to collect the corresponding warning image.

[0109] In an embodiment of the present application, a real-time image of a river flood control wall is collected; a corresponding first feature set is determined based on the real-time image of the river flood control wall and a recognition model; vertical settlement parameters and lateral displacement parameters are determined based on deformation detection of the river flood control wall by a GNSS displacement monitoring station; a corresponding warning data set is determined based on the vertical settlement parameters, lateral displacement parameters, and corresponding warning thresholds, and collection of a corresponding warning image is triggered based on the warning data set, which is compatible with multiple interactions of the vertical settlement parameters, lateral displacement parameters, and corresponding warning thresholds, thereby ensuring the accuracy of the warning data set and thus ensuring the accuracy of the warning image.

[0110] At this time, freeze the vertical settlement parameters and lateral displacement parameters; associate the vertical settlement parameters, lateral displacement parameters and the corresponding warning thresholds; compare the vertical settlement parameters, lateral displacement parameters and the corresponding warning thresholds in real time, thereby realizing real-time comparison of the vertical settlement parameters, lateral displacement parameters and the corresponding warning thresholds, and ensuring the control of the vertical settlement parameters, lateral displacement parameters and the corresponding warning thresholds.

[0111] At this time, the vertical settlement parameters and lateral displacement parameters collected in real time are associated with the preset warning thresholds. The warning threshold is determined based on the design requirements of the river flood control wall, historical deformation data, geological conditions, environmental factors, safety assessment standards and other factors, and is used to determine whether the deformation reaches or exceeds the dangerous level. In actual operation, this step is usually implemented through special monitoring software or systems. The monitoring software will preset an early warning threshold database, which stores the vertical settlement and lateral displacement early warning thresholds for different monitoring points. When the real-time monitoring system collects new deformation parameters, the software will automatically associate these parameters with the corresponding values ​​in the early warning threshold database for subsequent comparative analysis.

[0112] The associated vertical settlement parameters and lateral displacement parameters are compared with the corresponding warning thresholds in real time to determine whether the deformation has reached the warning state. This step is also achieved through monitoring software or systems. The software automatically compares the parameters and thresholds and gives corresponding prompts or alarms based on the comparison results. In actual operation, the comparison process needs to be fast and accurate so that a warning can be issued in time when the deformation reaches a dangerous level. At the same time, the software should also have the function of recording the comparison results and generating reports for subsequent analysis and processing.

[0113] Specifically, suppose that a river flood control wall has two monitoring points, A and B. The historical deformation data of point A is relatively stable, while the deformation of point B fluctuates greatly due to complex geological conditions. According to the design requirements and safety assessment standards, the vertical settlement warning threshold set for point A is 10mm, and the lateral displacement warning threshold is 5mm; the vertical settlement warning threshold set for point B is 5mm, and the lateral displacement warning threshold is 3mm. During a real-time monitoring process, the vertical settlement parameter of point A is 8mm, and the lateral displacement parameter is 4mm; the vertical settlement parameter of point B is 6mm, and the lateral displacement parameter is 2.5mm. The monitoring software associates these parameters with the corresponding values ​​in the warning threshold database. The results show that the deformation parameters of point A are all below the warning threshold, while the vertical settlement parameter of point B is close to the warning threshold.

[0114] The monitoring software has associated the deformation parameters of the two monitoring points A and B with the warning thresholds. In step S143, the software compares these parameters with the warning thresholds in real time.

[0115] The comparison results show that the vertical settlement parameter of point A, 8mm, and the lateral displacement parameter of point A, 4mm, are both below the warning threshold, so point A is in a safe state. The vertical settlement parameter of point B, 6mm, is close to the warning threshold of 5mm (although it has not exceeded it, it is close to the dangerous level), and the lateral displacement parameter of point B, 2.5mm, is below the warning threshold. The software gives corresponding prompts based on the comparison results, and recommends paying close attention to point B, strengthening the monitoring frequency and data analysis. At the same time, the software also generates a comparison report, which records in detail the deformation parameters, warning thresholds, and comparison results of the two monitoring points A and B for subsequent analysis and processing.

[0116] Furthermore, if the vertical settlement parameter and the lateral displacement parameter are greater than the corresponding warning threshold, the vertical settlement parameter and the lateral displacement parameter in the warning state are output; the vertical settlement parameter and the lateral displacement parameter in the warning state are used to determine the corresponding warning data set; the drone is triggered to detect again according to the warning data set, and the corresponding warning image is output to collect the corresponding warning image, which is compatible with the multiple interactions of the vertical settlement parameter, the lateral displacement parameter, and the corresponding warning threshold, ensuring the accuracy of the warning data set, thereby ensuring the accuracy of the warning image.

[0117] At this time, the monitoring system will check in real time whether the vertical settlement parameters and lateral displacement parameters of each monitoring point exceed the preset warning threshold. Once one or more parameters exceed the corresponding warning threshold, the system will immediately determine that the monitoring point is in a warning state and automatically output the vertical settlement parameters and lateral displacement parameters in this state. The output warning information usually includes the identification of the monitoring point, the time when the warning occurs, the specific value of the vertical settlement parameter, the specific value of the lateral displacement parameter, and the degree of exceeding the warning threshold. This information is crucial for subsequent analysis, decision-making and emergency response.

[0118] The monitoring system will integrate the vertical settlement parameters and lateral displacement parameters in the warning state into an early warning data set. This data set usually contains information such as the time period when the warning occurs, the identification of the monitoring point, the vertical settlement parameters and lateral displacement parameters at each time point. The formation of the early warning data set will help to conduct in-depth analysis of the early warning event in the future and understand the development trend and possible causes of the deformation. At the same time, the early warning data set is also an important basis for triggering the drone to detect again.

[0119] When the monitoring system generates a warning data set, it will automatically trigger the drone to conduct another inspection. The drone will fly to the area where the warning occurs, use the camera or other sensors on board to take warning images, and transmit the images back to the monitoring system. The warning images help to intuitively understand the actual situation in the warning area, including the specific location, scope and degree of deformation. This information is of great reference value for subsequent decision-making and emergency response.

[0120] Specifically, suppose a river flood control wall has a C monitoring point, the preset vertical settlement warning threshold is 15mm, and the lateral displacement warning threshold is 8mm. During a real-time monitoring process, the vertical settlement parameter of point C reached 16mm, and the lateral displacement parameter reached 9mm, both exceeding the corresponding warning thresholds.

[0121] The monitoring system immediately determined that point C was in a warning state and automatically output the warning information. The warning information included the identification of point C, the time of the warning (such as 14:30 on May 15, 2023), the vertical settlement parameter of 16mm, the lateral displacement parameter of 9mm, and the degree of exceeding the warning threshold (such as vertical settlement exceeding 1mm, lateral displacement exceeding 1mm).

[0122] The monitoring system has output the warning information of point C. In step S145, the system integrates the vertical settlement parameters and lateral displacement parameters of point C during the warning period into an early warning data set. The early warning data set contains information such as the identification of point C, the time period when the warning occurs (such as 14:30 to 14:45 on May 15, 2023), the vertical settlement parameters and lateral displacement parameters at each time point. For example, the data set may record that the vertical settlement parameter is 16mm and the lateral displacement parameter is 9mm at 14:30; the vertical settlement parameter is 16.5mm and the lateral displacement parameter is 9.2mm at 14:35, etc.

[0123] The system automatically triggered the drone to conduct another inspection based on the early warning data set. The drone flew to the area where point C was located and took early warning images using the high-definition camera on board. The image clearly showed that the river flood control wall at point C had obvious deformation, including cracks, tilting, and sinking. The drone transmitted the early warning images it took back to the monitoring system and stored them in a designated folder.

[0124] After receiving the warning image, the monitoring system automatically associates it with the warning data set and generates a complete warning report. This report contains warning information, warning data set, warning image and other information, providing comprehensive and accurate data support for subsequent analysis, decision-making and emergency response.

[0125] In addition, the drone automatically collects image data sets. The drone flight altitude, speed, camera focus and other parameter information are preset. According to the on-site environment, the flight altitude is specified to be 60 meters, the drone flight speed is 1m / s, and the drone camera focus is 20. To collect warning location information, the drone automatically points to fly according to the preset parameters. After arriving at the location, it automatically collects and forms an image data set. The data set covers the complete wall of the two sections of the flood control wall adjacent to the monitoring point, including front and top views, and 10 images with a pixel size of 1080*1920. The drone automatically transmits the data set back to the central server.

[0126] In step S15, a corresponding second feature set is determined based on the warning image and the recognition model, and a non-similar feature set is determined based on the comparison between the second feature set and the first feature set;

[0127] In the specific implementation process of the present invention, the specific steps may be:

[0128] S151: freeze warning image;

[0129] S152: Associating warning images and recognition models;

[0130] S153: Determine a corresponding second feature set according to the warning image and the recognition model;

[0131] S154: Compare the second feature set and the first feature set;

[0132] S155: Determine a non-similar feature set based on the comparison between the second feature set and the first feature set.

[0133] In an embodiment of the present application, a warning image is frozen; the warning image and the recognition model are associated; and a corresponding second feature set is determined based on the warning image and the recognition model, which is compatible with the overall consideration of the warning image and the recognition model, realizes multiple interactions of the warning image and the recognition model, and ensures the accuracy of the second feature set.

[0134] At this point, the frozen warning image is associated with a pre-trained recognition model. This recognition model is usually an image recognition model based on machine learning or deep learning, which can analyze the input image and extract key features or information. The association process involves inputting the warning image into the recognition model and ensuring that the model can correctly parse and process these images. This usually requires the recognition model to have sufficient generalization ability to handle images in different environments, lighting conditions, and shooting angles. At the same time, the association process may also involve some preprocessing steps, such as image enhancement, noise reduction, or resizing, to improve the accuracy and efficiency of the recognition model.

[0135] The recognition model is used to analyze the warning image and determine a set of corresponding features, called the second feature set. These features are usually key information extracted from the image and can describe the specific situation of the deformation, such as the width, length, shape of the crack, the angle of inclination, the range of sinking, etc.

[0136] The process of determining the second feature set involves output parsing and feature extraction of the recognition model. The recognition model usually performs multi-layer convolution and pooling operations on the input image to extract high-dimensional features in the image. These features are then further processed and parsed through a fully connected layer or classifier to obtain the final feature set.

[0137] Specifically, suppose that during a deformation monitoring of a river flood control wall, a drone took a warning image showing cracks and tilt in a certain part of the flood control wall. At this point, it is necessary to associate this warning image with a pre-trained image recognition model. The recognition model has been trained with a large amount of image data and can identify various deformation features of the river flood control wall, such as cracks, tilt, and sinking. In the association process, the warning image is first preprocessed, including enhancing image contrast and removing noise, to improve the accuracy of the recognition model. Then, the preprocessed image is input into the recognition model, which analyzes the image and extracts key features, such as the width, length, and tilt angle of the crack.

[0138] The warning image has been associated with the recognition model, and the analysis results of the recognition model on the image have been obtained. It is necessary to further analyze the output of the recognition model and determine the corresponding second feature set. The analysis results of the recognition model on the warning image show that there is an obvious crack in the image, the width of the crack is 5mm, the length is 2m, and the shape is approximately a straight line. At the same time, the recognition model also detected that the inclination angle of the flood control wall is 3 degrees. Based on this information, the second feature set can be determined as: {crack width: 5mm, crack length: 2m, crack shape: straight line, inclination angle: 3 degrees}. These features can accurately describe the deformation shown in the warning image, providing an important basis for subsequent analysis and decision-making.

[0139] Therefore, the second feature set and the first feature set are compared; based on the comparison between the second feature set and the first feature set, the non-similar feature set is determined, the comparison between the second feature set and the first feature set is achieved, and the accuracy of the non-similar feature set is ensured.

[0140] At this point, the second feature set extracted from the warning image is compared with the first feature set previously collected or calculated. The first feature set may come from a variety of sources, such as real-time monitoring data, historical image data, simulation results, etc., and it represents a certain "normal" or "expected" state. The comparison process involves comparing each feature in the two sets of feature sets one by one to evaluate the differences between them. This difference may be manifested as a difference in value, the presence or absence of a feature, or a change in the relative relationship between the features. The purpose of the comparison is to identify those features that appear in the second feature set but do not appear in the first feature set or are significantly different. These features may indicate potential problems or deformations.

[0141] According to the results of the comparison in step S154, a non-similar feature set is determined. This set contains those features that appear in the second feature set but are significantly different from the corresponding features in the first feature set. These non-similar features may represent key information in deformation monitoring, such as the expansion of cracks, the intensification of tilt, etc. The process of determining the non-similar feature set involves in-depth analysis and judgment of the comparison results. It is necessary to evaluate the importance and significance of each difference feature, and whether they are sufficient to trigger an early warning or take corresponding countermeasures. This usually requires a comprehensive judgment based on specific deformation monitoring needs and background knowledge.

[0142] Specifically, suppose that during a bridge deformation monitoring process, an early warning image is obtained by drone photography and

[0143] The second feature set was extracted, including information such as crack width, length and tilt angle of a certain part of the bridge. At the same time, the first feature set was obtained from the real-time monitoring data, representing the characteristic values ​​of the bridge in normal state. The second feature set was compared with the first feature set. It was found that the crack width increased from 2mm in normal state to 5mm, the crack length increased from 1m to 2m, and the tilt angle increased from 0.5 degrees to 3 degrees. These differences exceeded the preset threshold, indicating that a certain part of the bridge may have undergone serious deformation. Based on the comparison results, the non-similar feature set was determined as: {crack width increased by 3mm, crack length increased by 1m, and tilt angle increased by 2.5 degrees}. These non-similar feature sets accurately describe the specific situation of bridge deformation and provide an important basis for subsequent analysis and decision-making. Based on this information, relevant departments can quickly take measures to inspect and repair the bridge to ensure its safe operation.

[0144] In another embodiment of the present application, a BFMatcher matcher is used to match key points. The distance threshold in the matcher is set to 0.67. Only matches with distances below this threshold are accepted. Matches with distances above the threshold, or no matching key points are found at all, can be considered as non-similar feature sets in the image.

[0145] In step S16, multiple judgments are made on the non-similar feature set and the warning data set, and the recognition result of the non-similar feature set and the deformation result of the warning data set are output. The deformation result of the river flood control wall is determined according to the recognition result of the non-similar feature set, the deformation result of the warning data set and the result matching table;

[0146] In the specific implementation process of the present invention, the specific steps may be:

[0147] S161: Fixed non-similar feature set;

[0148] S162: Associate non-similar feature set and warning data set;

[0149] S163: Multiple judgments on non-similar feature sets and warning data sets;

[0150] S164: outputting the recognition result of the non-similar feature set and the deformation result of the warning data set based on the multiple judgments of the non-similar feature set and the warning data set;

[0151] S165: associating the recognition result of the non-similar feature set, the deformation result of the warning data set, and the result matching table;

[0152] S166: Determine river flood control based on the recognition results of the non-similar feature set, the deformation results of the warning data set, and the result matching table

[0153] The deformation results of the wall.

[0154] In the specific implementation process of the present invention, a corresponding second feature set is determined based on the early warning image and the recognition model, and a non-similar feature set is determined based on the comparison between the second feature set and the first feature set; multiple judgments are made on the non-similar feature set and the early warning data set, and the recognition results of the non-similar feature set and the deformation results of the early warning data set are output; the deformation results of the river flood control wall are determined according to the recognition results of the non-similar feature set, the deformation results of the early warning data set and the result matching table, and the recognition results of the non-similar feature set, the deformation results of the early warning data set and the result matching table are compatible, and multi-dimensional control of the recognition results of the non-similar feature set, the deformation results of the early warning data set and the result matching table is achieved, and multi-level control of the deformation results of the river flood control wall is ensured, thereby ensuring the accuracy of the deformation results of the river flood control wall, and realizing all-weather, timely, efficient and intelligent deformation detection of the river flood control wall, thereby improving the river risk prevention and control capabilities and providing strong guarantees for urban flood control safety.

[0155] At this time, freeze the non-similar feature set; associate the non-similar feature set and the warning data set; perform multiple judgments on the non-similar feature set and the warning data set, thereby ensuring multi-dimensional control of the non-similar feature set and the warning data set.

[0156] Therefore, based on the multiple judgments of the non-similar feature set and the warning data set, the recognition results of the non-similar feature set and the deformation results of the warning data set are output; at this time, the non-similar feature set is associated with the warning data set. The non-similar feature set is usually determined by comparing the first feature set and the second feature set in the previous step. These features reflect the abnormal conditions of the monitored object (such as the river flood control wall). The warning data set may contain a variety of data related to the deformation of the monitored object, such as real-time monitoring data, historical deformation records, expert evaluation results, etc. The purpose of association is to match and compare the non-similar feature set with the data in the warning data set, so as to have a more comprehensive understanding of the deformation of the monitored object. This association may involve multiple aspects such as data format conversion, timestamp alignment, and feature value matching.

[0157] The "increase in crack width" in the non-similar feature set is associated with the real-time monitoring crack width data in the early warning data set, and the "increase in wall inclination angle" is associated with the real-time monitoring wall inclination angle data. At the same time, historical deformation records are also consulted to understand whether these non-similar features have appeared in the past and the deformation conditions at that time. Multiple judgments are made on the non-similar feature set and the early warning data set. Multiple judgments mean analyzing and evaluating these two sets of data from multiple angles, using multiple methods and indicators. This helps to have a more comprehensive understanding of the deformation of the monitored object and reduce the errors or biases that may be caused by a single judgment method.

[0158] Evaluate the correlation and change trend between the non-similar feature sets and the warning data set. Classify or identify the non-similar feature sets to understand the type or degree of deformation they may represent. Combine expert experience and domain knowledge to interpret and evaluate the analysis results to determine the deformation and potential risks of the monitored object.

[0159] Based on the multiple judgment results of the non-similar feature set and the warning data set, output the recognition results of the non-similar feature set and the deformation results of the warning data set. The recognition results may include a specific description, classification or explanation of the non-similar features, while the deformation results may include information such as the degree of deformation, trend or potential risk of the monitored object. Organize and analyze the multiple judgment results to extract key information and conclusions. Write a report or explanatory document of the recognition results and deformation results to ensure that the information is accurate, clear and easy to understand. Communicate the recognition results and deformation results to relevant personnel or departments in a timely manner so that they can take appropriate countermeasures.

[0160] Specifically, suppose that in a certain river flood control wall deformation monitoring, the non-similar feature set includes two features: "increased crack width" and "increased wall inclination angle". The early warning data set contains the real-time monitored crack width data, wall inclination angle data, and historical deformation records. The "increased crack width" in the non-similar feature set is associated with the real-time monitored crack width data in the early warning data set, and the "increased wall inclination angle" is associated with the real-time monitored wall inclination angle data. At the same time, the historical deformation records are also consulted to understand whether these non-similar features have appeared in the past and the deformation conditions at that time.

[0161] First, the trend analysis method was used to evaluate the changing trends of crack width and wall inclination angle, and it was found that they both showed a gradual increasing trend. Then, the cluster analysis algorithm was used to classify the non-similar feature sets, and it was found that these features could be clustered into two categories: "wall crack expansion" and "wall inclination intensification". Finally, combined with expert experience and domain knowledge, it was determined that these deformations may represent structural safety hazards of the river flood control wall, and immediate measures need to be taken to reinforce and repair it.

[0162] Based on the results of multiple judgments, the following recognition results and deformation results were output: Recognition results: The non-similar feature set was identified as two types of features, "wall crack expansion" and "wall inclination intensification", which respectively represent two different deformation conditions of the river flood control wall.

[0163] Deformation results: The river flood control wall has moderate deformation, with cracks in the wall gradually expanding and the wall inclination angle gradually increasing. Based on the current deformation trend and expert evaluation results, it is believed that the river flood control wall has structural safety hazards and immediate measures need to be taken to strengthen and repair it. These results were promptly communicated to relevant departments and personnel so that they can take corresponding countermeasures.

[0164] Furthermore, the recognition results of non-similar feature sets, the deformation results of the early warning data set and the result matching table are associated; the deformation results of the river flood control wall are determined based on the recognition results of the non-similar feature sets, the deformation results of the early warning data set and the result matching table, which are compatible with the recognition results of non-similar feature sets, the deformation results of the early warning data set and the result matching table, and realize multi-dimensional control of the recognition results of non-similar feature sets, the deformation results of the early warning data set and the result matching table, thereby ensuring multi-level management and control of the deformation results of the river flood control wall, and then ensuring the accuracy of the deformation results of the river flood control wall, and realizing all-weather, timely, efficient and intelligent deformation detection of the river flood control wall, thereby improving the river risk prevention and control capabilities, and providing strong guarantees for urban flood control safety.

[0165] At this point, the recognition results of the non-similar feature set, the deformation results of the warning data set, and the result matching table are associated. The result matching table is usually a pre-defined table or database that contains the correspondence between different deformation features and warning levels. The purpose of this association is to compare the recognition results and deformation results with the result matching table to determine the specific deformation level or warning status of the monitored object (such as a river flood control wall).

[0166] According to the results of the association in step S165, the final deformation result of the monitored object (such as the river flood control wall) is determined. This deformation result may be a specific value, level or state description, which combines the recognition results of the non-similar feature set, the deformation results of the warning data set, and the information of the result matching table. This deformation result will be used for subsequent decision-making and the adoption of response measures.

[0167] Specifically, assume that in a certain river flood control wall deformation monitoring project, steps S161 to S164 have been completed, and the recognition results of the non-similar feature set and the deformation results of the warning data set have been obtained. The recognition results include two features, "wall crack expansion" and "wall inclination intensification", and the deformation results indicate that the river flood control wall has a moderate deformation trend. In step S165, the result matching table is consulted. The table lists the correspondence between different deformation features and warning levels, for example, "wall crack expansion" may correspond to the "yellow warning" level, and "wall inclination intensification" may correspond to the "orange warning" level. At the same time, the table also takes into account the degree of deformation, such as moderate deformation may correspond to a higher warning level.

[0168] After comparing the recognition results and deformation results with the result matching table, it was found that the two features of "wall crack expansion" and "wall tilt intensification" both reached the "orange warning" level. Therefore, the deformation level of the river flood control wall was determined to be "orange warning".

[0169] According to the results of the association in step S165, the deformation result of the river flood control wall is determined to be "orange warning". This result shows that the river flood control wall has an obvious deformation trend and immediate measures need to be taken to strengthen and repair it.

[0170] After the deformation results were determined, relevant personnel immediately took corresponding countermeasures, such as increasing monitoring frequency, strengthening inspections, and formulating reinforcement plans. These measures are aimed at ensuring the safe operation of the river flood control wall and preventing potential risks caused by deformation. At the same time, the deformation results are also recorded in the project report for subsequent analysis and reference.

[0171] In addition, joint judgment is carried out through non-similar feature sets and early warning data sets. The details are as follows:

[0172] (1) If the GNSS displacement monitoring station determines that there is deformation, but the intelligent vision model detection results show that there is no deformation,

[0173] The possible reason is that the GNSS displacement monitoring station is affected by heavy trucks passing nearby, resulting in large signal fluctuations, or electromagnetic interference, which leads to false alarms.

[0174] (2) If the GNSS displacement monitoring station determines that there is deformation, and the intelligent visual model detection results show an abnormality, it is determined that deformation has occurred. The central server will notify the on-duty personnel of the relevant units via text messages, smart phones, etc., so that timely response measures can be taken.

[0175] In an embodiment of the present invention, through the method in the embodiment of the present invention, real-time images of river flood control walls are collected; the corresponding first feature set is determined according to the real-time images of the river flood control walls and the recognition model; the vertical settlement parameters and lateral displacement parameters are determined based on the deformation detection of the river flood control wall by the GNSS displacement monitoring station; the corresponding warning data set is determined according to the vertical settlement parameters, the lateral displacement parameters, and the corresponding warning thresholds, and the collection of the corresponding warning image is triggered according to the warning data set, which is compatible with multiple interactions of the vertical settlement parameters, the lateral displacement parameters, and the corresponding warning thresholds, thereby ensuring the accuracy of the warning data set and thus ensuring the accuracy of the warning image.

[0176] Furthermore, a corresponding second feature set is determined based on the warning image and the recognition model, and a non-similar feature set is determined based on the comparison between the second feature set and the first feature set; multiple judgments are made on the non-similar feature set and the warning data set, and the recognition results of the non-similar feature set and the deformation results of the warning data set are output; the deformation results of the river flood control wall are determined based on the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table, which are compatible with the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table, and realize multi-dimensional control of the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table, ensuring multi-level control of the deformation results of the river flood control wall, and then ensuring the accuracy of the deformation results of the river flood control wall, realizing all-weather, timely, efficient and intelligent deformation detection of the river flood control wall, thereby improving the river risk prevention and control capabilities and providing strong guarantees for urban flood control safety.

[0177] Please refer to FIG. 2 , which is a schematic diagram of the structural composition of a deformation detection system for a river flood control wall in an embodiment of the present invention.

[0178] As shown in FIG. 2 , a deformation detection system for a river flood control wall is provided, and the deformation detection system for a river flood control wall comprises:

[0179] The acquisition module 21 is used to acquire real-time images of the river flood control wall;

[0180] A first feature module 22, used to determine a corresponding first feature set according to the real-time image of the river flood control wall and the recognition model;

[0181] The deformation detection module 23 is used to determine the vertical settlement parameters and the lateral displacement parameters based on the deformation detection of the river flood control wall by the GNSS displacement monitoring station;

[0182] The early warning module 24 is used to determine a corresponding early warning data set according to the vertical settlement parameter, the lateral displacement parameter, and the corresponding early warning threshold, and trigger the collection of the corresponding early warning image according to the early warning data set;

[0183] A comparison module 25, configured to determine a corresponding second feature set based on the warning image and the recognition model, and determine a non-similar feature set based on a comparison between the second feature set and the first feature set;

[0184] The result matching module 26 is used to perform multiple judgments on the non-similar feature set and the warning data set, and output the recognition results of the non-similar feature set and the deformation results of the warning data set, and determine the deformation results of the river flood control wall according to the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table.

[0185] Please refer to Figure 3. The electronic device 40 according to this embodiment of the present invention is described below with reference to Figure 3. The electronic device 40 shown in Figure 3 is only an example and should not bring any limitation to the function and scope of use of the embodiment of the present invention.

[0186] As shown in Fig. 3, the electronic device 40 is in the form of a general computing device. The components of the electronic device 40 may include but are not limited to: at least one processing unit 41, at least one storage unit 42, and a bus 43 connecting different system components (including the storage unit 42 and the processing unit 41).

[0187] The storage unit stores program codes, which can be executed by the processing unit 41, so that the processing unit 41 executes the steps according to various exemplary embodiments of the present invention described in the above “Embodiment Method” section of this specification.

[0188] The storage unit 42 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 421 and / or a cache memory unit 422 , and may further include a read-only memory unit (ROM) 423 .

[0189] The storage unit 42 may also include a program / utility 424 having a set (at least one) of program modules 425, such program modules 425 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0190] Bus 43 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0191] The electronic device 40 may also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 40, and / or may communicate with any device that enables the electronic device 40 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 44. Furthermore, the electronic device 40 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 45. As shown in FIG. 3 , the network adapter 45 communicates with other modules of the electronic device 40 via a bus 43. It should be understood that, although not shown in FIG. 3 , other hardware and / or software modules may be used in conjunction with the electronic device 40, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup planning systems, etc.

[0192] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (which can be a personal computer, a server, a multi-parameter sensor device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0193] Those skilled in the art can understand that all or some of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium can include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc. In addition, it stores computer program instructions, and when the computer program instructions are executed by a computer, the computer executes the above method.

[0194] In addition, the deformation detection method and system of the river flood control wall provided by the embodiment of the present invention are introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for detecting deformation of a river flood control wall, characterized in that: Applied to deformation detection scenarios of river flood control walls; The deformation detection method of the river flood control wall comprises: Collect real-time images of river flood control walls; Determine a corresponding first feature set according to the real-time image of the river flood control wall and the recognition model; Determine the vertical settlement parameters and lateral displacement parameters based on the deformation detection of the river flood control wall by the GNSS displacement monitoring station; Determine a corresponding warning data set according to the vertical settlement parameter, the lateral displacement parameter, and the corresponding warning threshold, and trigger the collection of the corresponding warning image according to the warning data set; Determine a corresponding second feature set based on the warning image and the recognition model, and determine a non-similar feature set based on a comparison between the second feature set and the first feature set; Multiple judgments are made on the non-similar feature set and the early warning data set, and the recognition results of the non-similar feature set and the deformation results of the early warning data set are output. The deformation results of the river flood control wall are determined according to the recognition results of the non-similar feature set, the deformation results of the early warning data set and the result matching table.

2. The deformation detection method of a river flood control wall according to claim 1 is characterized in that: The real-time image collection of the river flood control wall includes: Collect the location of the river flood control wall; Determine the corresponding flight detection area based on the location of the river flood control wall; Determine the flight path of the drone based on the environmental parameters of the flight detection area and the river flood control wall; The dynamic flight of the drone is triggered based on the flight path of the drone, and the drone takes dynamic photos of the river flood control wall during the dynamic flight; Real-time images of the river flood control wall are collected based on the dynamic photography of the river flood control wall by the drone.

3. The deformation detection method of a river flood control wall according to claim 2 is characterized in that: The determining of the corresponding first feature set according to the real-time image of the river flood control wall and the recognition model includes: Freeze real-time images of river flood control walls; Determine the grayscale image based on the real-time image of the river flood control wall and the perceptual weighted grayscale algorithm; According to the grayscale image and Gaussian blur algorithm, the corresponding noise reduction is triggered, and the image after noise reduction is output; Correlate the denoised image and the recognition model; A corresponding first feature set is determined according to the image after denoising and the recognition model.

4. The deformation detection method of a river flood control wall according to claim 3 is characterized in that: The vertical settlement parameters and lateral displacement parameters are determined based on the deformation detection of the river flood control wall by the GNSS displacement monitoring station, including: Fixed-frame GNSS displacement monitoring station; The GNSS displacement monitoring station detects deformation of the river flood control wall; Real-time monitoring of the deformation detection of river flood control walls involves data collection, processing and analysis to obtain deformation parameters: the collected data will be transmitted to the data center or monitoring center in real time for further processing and analysis. By analyzing the processed data, the deformation parameters of the river flood control wall can be obtained, including vertical settlement amount and rate, lateral displacement amount and direction; In the deformation detection of river flood control walls, the detection objects of the GNSS displacement monitoring station are collected. The detection objects usually refer to the key parts or the overall structure of the river flood control wall; The vertical settlement parameters and lateral displacement parameters are determined based on the real-time detection of the detection object and the river flood control wall.

5. The deformation detection method of a river flood control wall according to claim 4 is characterized in that: The method of determining a corresponding warning data set according to the vertical settlement parameter, the lateral displacement parameter, and the corresponding warning threshold, and triggering the collection of a corresponding warning image according to the warning data set, includes: Fixed vertical settlement parameters and lateral displacement parameters; Associate vertical settlement parameters, lateral displacement parameters, and corresponding warning thresholds; Compare vertical settlement parameters and lateral displacement parameters with corresponding warning thresholds in real time; If the vertical settlement parameter and the lateral displacement parameter are greater than the corresponding warning threshold, the vertical settlement parameter and the lateral displacement parameter in the warning state are output; Determine the corresponding warning data set based on the vertical settlement parameters and lateral displacement parameters in the warning state; According to the early warning data set, the drone is triggered to detect again, and the corresponding early warning image is output to collect the corresponding early warning image.

6. The deformation detection method of a river flood control wall according to claim 5, characterized in that: The determining of the corresponding second feature set based on the warning image and the recognition model, and determining the non-similar feature set based on the comparison between the second feature set and the first feature set, includes: Freeze warning images; Correlate warning images and recognition models; A corresponding second feature set is determined according to the warning image and the recognition model.

7. The deformation detection method of a river flood control wall according to claim 6, characterized in that: The method of determining the corresponding second feature set based on the warning image and the recognition model, and determining the non-similar feature set based on the comparison between the second feature set and the first feature set, further includes: Compare the second feature set and the first feature set; A non-similar feature set is determined based on a comparison between the second feature set and the first feature set.

8. The deformation detection method of a river flood control wall according to claim 7, characterized in that: The method of performing multiple judgments on the non-similar feature set and the early warning data set, and outputting the recognition result of the non-similar feature set and the deformation result of the early warning data set, and determining the deformation result of the river flood control wall according to the recognition result of the non-similar feature set, the deformation result of the early warning data set and the result matching table, includes: Fix non-similar feature sets; Associating non-similar feature sets and warning data sets; Make multiple judgments on non-similar feature sets and warning data sets.

9. The deformation detection method of a river flood control wall according to claim 8, characterized in that: The method further comprises: performing multiple judgments on the non-similar feature set and the early warning data set, and outputting the recognition result of the non-similar feature set and the deformation result of the early warning data set, and determining the deformation result of the river flood control wall according to the recognition result of the non-similar feature set, the deformation result of the early warning data set and the result matching table. Based on multiple judgments of the non-similar feature set and the warning data set, the recognition result of the non-similar feature set and the deformation result of the warning data set are output; Associate the recognition results of the non-similar feature set, the deformation results of the warning data set, and the result matching table; The deformation results of the river flood control wall are determined based on the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table.

10. A deformation detection system for a river flood control wall, characterized in that: The deformation detection system of the river flood control wall is applied to the deformation detection method of the river flood control wall as claimed in any one of claims 1 to 9, and the deformation detection system of the river flood control wall comprises: The acquisition module is used to collect real-time images of the river flood control wall; A first feature module, used to determine a corresponding first feature set according to a real-time image of a river flood control wall and a recognition model; The deformation detection module is used to determine the vertical settlement parameters and lateral displacement parameters based on the deformation detection of the river flood control wall by the GNSS displacement monitoring station; An early warning module is used to determine a corresponding early warning data set according to the vertical settlement parameter, the lateral displacement parameter, and the corresponding early warning threshold, and trigger the collection of a corresponding early warning image according to the early warning data set; A comparison module, used to determine a corresponding second feature set based on the warning image and the recognition model, and determine a non-similar feature set based on a comparison between the second feature set and the first feature set; The result matching module is used to make multiple judgments on the non-similar feature set and the warning data set, and output the recognition results of the non-similar feature set and the deformation results of the warning data set. The deformation results of the river flood control wall are determined according to the recognition results of the non-similar feature set, the deformation results of the warning data set and the result matching table.

Images