Visual measurement and early warning method and device for seawall overtopping height

Abstract

The invention discloses a seawall overtopping height visual measurement and early warning method and a seawall overtopping height visual measurement and early warning device. The invention aims to provide a method and a matched device for measuring the continuous height of seawall overtopping waves in a non-contact manner by adopting a visual means and carrying out early warning on overtopping waves beyond warning. The system has the functions of seawall overtopping visual monitoring, spindrift height real-time calculation and breakwater overtopping judgment, and seawall submerging pre-alarm is achieved. The method comprises the following steps: S1, obtaining seawall height measurement data, and generating a virtual equal-height field plane; s2, carrying out seawall overtopping visual monitoring to obtain a spindrift coverage area; s3, combining the virtual equal-height field plane and the spindrift coverage area to calculate the spindrift height; and S4, triggering a seawall submerging pre-alarm when the height exceeds the warning spray height. According to the invention, the seawall safety can be evaluated in real time, the seawall safety pre-alarm is realized, the alarm accuracy is high, and the seawall emergency response system is helped to be perfected.

Description

Technical Field

[0001] The present invention relates to the field of marine disaster prevention and mitigation in water conservancy projects, and in particular to a method and device for real-time assessment of seawall safety under disasters such as storm surges. Background Art

[0002] Marine disasters pose a major threat to coastal areas, among which the problems caused by storm surges are increasingly becoming the main challenge facing coastal cities. Every summer, coastal areas are often hit by typhoons, resulting in disasters such as seawater flooding and inundation caused by storm surges, posing a serious threat to people's lives, property and the ecological environment. Storm surges are one of the main marine disasters faced by coastal areas, which can easily cause disasters such as seawater flooding and inundation. Among them, overtopping, as an important indicator, describes the flow of water over the top of coastal structures such as seawalls under the action of waves. Its size is directly related to the stability of the seawall structure and the safety of coastal areas.

[0003] However, for a long time, obtaining the overtopping amount of seawalls in nearshore areas has been a difficult point in nearshore observation work. Most of the previous attempts were based on laboratory experiments or empirical formulas. These methods have limitations and shortcomings: although the measurements are accurate under laboratory conditions, they are difficult to reflect the actual process, and the scope of application of the existing formulas is limited, resulting in large differences in the calculation results. Even if there are studies based on numerical simulations, their verification lacks sufficient basic data support. These limitations make it difficult to accurately grasp the characteristics of storm surge overtopping flooding disasters, so reliable, continuous, and long-term measured data of the overtopping process are urgently needed. In addition, the field monitoring technology of seawall overtopping is usually based on sensors, which are often susceptible to severe weather conditions, resulting in inaccurate and unreliable monitoring data. Traditional seawall overtopping monitoring technology usually requires dense installation of sensors, so it is impossible to achieve comprehensive and continuous monitoring of seawalls, which limits the coverage and real-time nature of monitoring. In addition, there is also a method that uses active measurement methods such as lidar and infrared light to monitor the overtopping of seawalls. However, since the water body absorbs, refracts, and transmits laser and infrared light, the active light signal will weaken sharply, which will lead to the failure of seawall overtopping measurement. Therefore, this type of method has extremely strict restrictions on scenes and environments and does not have the conditions for general on-site implementation. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention proposes a method and device for visual measurement and early warning of seawall overtopping height. The method and device of the present invention use visual means to measure the height of seawall overtopping without contact, and can issue early warnings for overtopping that exceeds the warning level. Seawall inundation early warning is achieved through visual monitoring of seawall overtopping, real-time calculation of wave height, and judgment of overtopping. This new method and device can achieve real-time assessment of seawall safety, early warning of seawall safety, and high alarm accuracy, thereby providing strong support for improving the seawall emergency response system.

[0005] To achieve the above object, the present invention adopts the following technical solutions: A method for visual measurement and early warning of seawall overtopping height, comprising the following steps: S1: Measure the height information of different positions of the seawall at the seawall site, and generate a virtual contour field plane of the seawall site by combining the contour field interpolation algorithm; S2: Visually monitor the seawall and use the wave semantic segmentation model to perform semantic segmentation and recognition on the images of seawall overtopping waves to obtain the wave coverage area; S3: Combining the virtual contour field plane of the seawall site and the semantic segmentation wave coverage, the edge extraction algorithm is used to extract the wave edge corresponding to the semantic segmentation wave coverage, and then the contour field coordinate mapping algorithm is used to map the wave edge coordinates to the virtual contour field plane of the seawall site to obtain continuous wave height; S4: If any of the continuous wave heights exceeds the warning level, the seawall flooding warning alarm will be triggered.

[0006] Furthermore, the method for measuring the height information of different positions of the seawall at a fixed point on the seawall site in step S1, the specific measurement method includes the following steps: Sa1: Use a height measuring rod device to measure the fixed point height of the seawall, and record the physical length unit reading when the height measuring rod is used to measure the fixed point height; Sa2: Take the image when the height measuring rod is used to measure the height, and record the image pixel unit reading when the height measuring rod device is used to measure the height at a fixed point; Sa3: Establish the corresponding relationship between the physical length unit reading and the image pixel unit reading when the height measuring rod device is used for fixed-point height measurement, and complete the process of measuring the height information of the seawall at a fixed point; Sa4: Repeat steps Sa1 to Sa3, perform the process of measuring the height information of the seawall at least 3 times, and establish the corresponding relationship between the physical length unit reading and the image pixel unit reading when measuring the height at a fixed point using at least 3 height measuring rod devices.

[0007] Furthermore, the wave semantic segmentation model in step S2 includes the following modules: Data collection module: collects seawall overtopping image data for training and evaluation, and annotates the collected seawall overtopping images; Model training module: used to train the semantic segmentation model so that it can accurately identify the wave-covered area in the seawall overtopping image; Model recognition module: used to perform semantic segmentation prediction on the collected seawall overtopping images and identify the wave coverage area.

[0008] Furthermore, the data acquisition module collects seawall overtopping image sets under different weather and sea conditions. And annotated dataset ; The seawall overtopping image set And annotated dataset The collection process covers various sea conditions and lighting conditions, as follows: Factor 1: Collection frequency: The collection frequency is set to once an hour to ensure timely monitoring and early warning of the seawall overtopping status; Factor 2: Weather factors: When collecting data, the impact of different weather conditions is taken into account, including sunny, cloudy, rainy, etc., to ensure that data collection is all-weather to meet the needs of seawall overtopping monitoring under different weather conditions; Factor 3: Day and night factors: When collecting data, the changes of day and night should be taken into account. The camera should have the function of night shooting. At night, additional lighting devices should be used to ensure image quality to ensure all-weather monitoring of seawall overtopping; Factor 4: Data desensitization and privacy processing: During the data collection process, data desensitization should be performed for parts involving personal privacy to ensure privacy security. The collection, storage and transmission of all data must comply with relevant privacy protection laws and standards; Factor 5: Data diversity: Data collection should cover different locations and angles of the seawall to ensure the robustness and generalization ability of the model. In addition, images under different lighting conditions should be collected to improve the adaptability of the model; Factor 6: Marking quality control: When marking the wave coverage area, take measures such as multi-person marking and marking result review to ensure the accuracy and consistency of the marking; Factor 7: Sea condition monitoring: During the data collection process, sea conditions, including wind speed, wind direction, tide and other information, are recorded for future analysis and model improvement; Factor 8: Data storage: The collected seawall overtopping image data is stored in a secure and reliable database for subsequent use in training and validating models; Factor 9: Data collection site setting: Select representative seawall overtopping monitoring sites for data collection, covering seawall overtopping conditions under different geographical conditions to ensure the universality and reliability of the model; Furthermore, the implementation process of the model training module includes the following specific steps: S521: Data preprocessing: The seawall overtopping image set And annotated dataset Preprocessing is performed, and the preprocessing methods include: image grayscale, image rotation, image reduction, image enlargement, image cropping, image displacement, histogram equalization, and high-pass filter kernel smoothing; S522: Model initialization: The CNN parameters of each level of the convolutional neural network of the wave semantic segmentation model are initialized using the Xaiver parameter initialization algorithm. The Xaiver parameter initialization algorithm is mathematically expressed as: (1) Where: Indicates from a arrive b Uniform distribution of is the number of input connections of the previous layer of the convolutional neural network; is the number of output connections of the last layer of the convolutional neural network; S523: Model training: Using pre-processed seawall overtopping image set And annotated dataset , the spray semantic segmentation model after model initialization is trained, and the model training loss function is the cross entropy loss function, which is mathematically expressed as: (2) Where: N The seawall overtopping image set And annotated dataset The number of samples; is the labeled dataset The actual annotations corresponding to the samples in The seawall overtopping image set The output obtained after the image corresponding to the sample in is input into the Langhua semantic segmentation model.

[0009] Furthermore, the implementation process of the model identification module includes the following specific steps: S531: Data preprocessing: image preprocessing is performed on the collected seawall overtopping images, including image grayscale processing, histogram equalization, and high-pass filter kernel smoothing; S532: Model recognition: Use the pre-processed seawall overtopping image to input the wave semantic segmentation model to obtain the wave coverage mask , in the waves covered with masks Any pixel in m Its mathematical expression is: (3) Where: It is the wave-covered mask; m It's a wave-covered mask A pixel in .

[0010] Furthermore, the edge extraction algorithm in step S3 includes the following specific steps: S31: Initialization: Start from the upper left corner edge point of the image and traverse until the wave cover mask is found The first one to satisfy m= The pixel point of 1 is used as the starting point for edge extraction; S32: Record the current pixel: mark the current pixel as visited and record the coordinates of the current pixel ; S33: Go to the next pixel: Check the eight neighboring pixels in the up, down, left, right, and four diagonal directions of the current pixel in a clockwise or counterclockwise order, and select one neighboring pixel that has not been visited and satisfies m= The pixel with value 1 is taken as the next edge extraction point and marked as visited; S34: Repeat step S33: Loop through step S33 until you return to the starting point or find no accessible pixels; if there are multiple unvisited pixels in the pixel neighborhood that satisfy m= 1, select the pixel farthest from the current pixel to ensure that the contour is traced as long as possible; S35: Termination condition: The algorithm terminates after returning to the starting point, or when there are no more unvisited nodes and the m= The pixel is 1 and ends.

[0011] Furthermore, the mathematical expression of the contour field coordinate mapping algorithm in step S3 is: (4) Where: is the coordinate of the wave edge extracted by the edge extraction algorithm in the image pixel unit; is the coordinate of the wave edge extracted by the edge extraction algorithm in physical length units; and The virtual contour plane of the seawall site is x Axis and y The result of isovalue field interpolation on the axis.

[0012] Furthermore, the triggering of the seawall flooding pre-alarm in step S4 is specifically carried out in the following manner: when the wave height at any point in the continuous wave heights obtained by visual measurement is greater than h Greater than the warning height , then the seawall flooding warning alarm is triggered, and its mathematical expression is: (5) Where: H is the continuous wave height obtained by visual measurement; h is the wave height at one point in the continuous wave height; is the pre-set alert altitude, and The value of is in [0m, 30m].

[0013] The present invention provides an integrated device for visual monitoring of seawall overtopping that is compatible with the method of the present invention, comprising a camera lens, a camera, a metal pole, a terminal equipment box for visual monitoring of seawall overtopping, a supporting connection structure for connecting the camera and the metal pole and playing a supporting role, a supporting connection structure for connecting the terminal equipment box for visual monitoring of seawall overtopping and the metal pole and playing a supporting role, and a pole base for fixing the metal pole.

[0014] Furthermore, the seawall wave overrun visual monitoring terminal equipment box includes a terminal equipment box door, a terminal equipment box lock, a terminal equipment box body, a hinge for connecting the terminal equipment box door and the terminal equipment box body, a power supply socket strip, a terminal equipment network switch, a wireless communication network card, and a terminal equipment server. The seawall wave overrun visual monitoring terminal equipment box is connected to a metal vertical pole through a supporting connection structure.

[0015] Furthermore, the terminal device server includes a processor and a memory; the memory stores at least one program, and the at least one program is loaded and executed by the processor to implement the method for visual measurement and early warning of seawall overtopping height.

[0016] Furthermore, the terminal device server includes a computer-readable storage medium; the computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for loading by a processor to execute the method for visual measurement and early warning of seawall overtopping height.

[0017] The beneficial effects of the present invention are: 1. The present invention adopts a vision-based seawall overtopping height measurement method, which has a wider monitoring range than the single-point sensor method, and can adapt to the characteristics of the water body, has lower environmental requirements, and is more robust than the active light monitoring methods of laser and infrared. Through visual monitoring of seawall overtopping, comprehensive and continuous monitoring of the seawall overtopping status can be achieved, which is not limited by the number of sensors and the layout location. Compared with the traditional single-point sensor method, the vision-based continuous measurement method has a wider monitoring range and can achieve more comprehensive and accurate monitoring of the seawall overtopping situation. At the same time, compared with active light monitoring methods such as laser and infrared, the vision-based method does not need to emit light, so it has lower environmental requirements, can adapt to a variety of weather and lighting conditions, and has stronger robustness.

[0018] 2. The present invention realizes an integrated device for visual monitoring of seawall overtopping that is compatible with the method, which integrates cameras, terminal equipment servers, wireless communication network cards and other equipment, and has a simple structure, low cost, reasonable design and easy operation. The integrated device can be directly installed on the seawall without the need for a complicated installation process, and does not occupy a large amount of space on the seawall. The device integrates cameras, servers and other equipment together, greatly simplifying the construction of the seawall overtopping monitoring system, reducing the installation, maintenance and operating costs of the system, and improving the stability and reliability of the system. Through the use of the integrated device, the visual measurement and early warning of the seawall overtopping height become simpler and more efficient.

[0019] 3. The present invention proposes a method for visual measurement and early warning of seawall overtopping height, which combines artificial intelligence algorithm recognition with hardware equipment to achieve unmanned monitoring of the seawall site around the clock. It has the characteristics of intelligence, real-time, and easy migration, and provides strong support for improving the seawall emergency response system. Through the recognition of artificial intelligence algorithms, real-time analysis of seawall overtopping images can be achieved, and early warning of over-warning wave heights can be issued without human intervention. At the same time, the vision-based continuous measurement method combined with artificial intelligence algorithms realizes unmanned monitoring of the seawall site around the clock, greatly improving the real-time and accuracy of monitoring, and providing more effective technical support for the safety management of seawalls. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of the steps of the method of the present invention; Figure 2 A flow chart of the steps of measuring the height of a seawall at a fixed point according to the present invention; Figure 3 It is a structural schematic diagram of the device of the present invention; Figure 4 It is a structural schematic diagram of a terminal equipment box for visual monitoring of seawall overtopping of the present invention; Figure 5 A schematic diagram of module calling of the wave semantic segmentation model of the present invention; Figure 6a A schematic diagram of a side view scene of the implementation site of the seawall overtopping monitoring of the present invention; Figure 6b A schematic diagram of a bird's-eye view of the implementation site of the seawall overtopping monitoring of the present invention; Figure 6c A schematic diagram of a front view scene of the seawall overtopping monitoring implementation site of the present invention; Figure 7 It is a diagram showing an implementation example of the integrated device for visual monitoring of seawall overtopping of the present invention; Figure 8 This is a diagram showing an implementation example of a terminal equipment box for visual monitoring of seawall overtopping according to the present invention; Fig. 9 It is an implementation example diagram of the virtual contour field plane of the seawall site of the present invention; Fig.10a It is a scene example diagram of the no overtopping no warning identification result of the present invention; Fig.10b It is a scene example diagram of the identification result of overtopping without warning of the present invention; Fig.10c It is a scene example diagram of the overtopping and early warning identification result of the present invention; Fig.11 It is a training curve diagram of the model training process of the model training module of the present invention.

[0021] In the figure: 510-data acquisition module; 520-model training module; 530-model recognition module; 301-camera lens; 302-camera; 303-metal pole; 304-seawall overtopping visual monitoring terminal equipment box; 305-first supporting connection structure; 306-second supporting connection structure; 307-pole base; 401-terminal equipment box door; 402-terminal equipment box lock; 403-terminal equipment box body; 404-hinge; 405-power supply socket; 406-terminal equipment network switch; 407-wireless communication network card; 408-terminal equipment server; 601-seawall fixed-point measurement height point; 602-seawall on-site virtual contour field plane; 603-camera visual range; 604-seawall overtopping visual monitoring integrated device; 605-seawall platform; 606-sea surface; 607-seawall platform to be observed. DETAILED DESCRIPTION

[0022] The division of modules in the present invention is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between modules can be electrical or other similar forms, which are not limited in the present invention. In addition, the modules or submodules described as separate components may or may not be physically separated, may or may not be physical modules, or may be distributed in multiple circuit modules, and some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention.

[0023] In the present invention, the term "and / or" is intended to cover all possible combinations and subcombinations of the listed elements, including any one, any subcombination or all of the elements listed separately, without excluding other elements. Unless otherwise specified, the directions or positional relationships indicated by the terms "front, back, top, bottom, left, right" and the like are usually based on the directions or positional relationships shown in the drawings, and are only for the convenience and simplification of description, and cannot be understood as limiting the scope of protection of the present invention.

[0024] The present invention is further described in detail below in conjunction with embodiments and drawings. Example 1

[0025] Example 1 focuses on the implementation steps and parameter setting details of the method provided by the present invention. Figure 1 , Figure 2 and Figure 5 As shown, the present invention provides a method for visual measurement and early warning of seawall overtopping height. For specific method flow, please refer to Figure 1 , including the following steps: S1: Measure the height information of different positions of the seawall at the seawall site, and generate a virtual contour field plane of the seawall site by combining the contour field interpolation algorithm; S2: Visually monitor the seawall and use the wave semantic segmentation model to perform semantic segmentation and recognition on the images of seawall overtopping waves to obtain the wave coverage area; S3: Combining the virtual contour field plane of the seawall site and the semantic segmentation wave coverage, the edge extraction algorithm is used to extract the wave edge corresponding to the semantic segmentation wave coverage, and then the contour field coordinate mapping algorithm is used to map the wave edge coordinates to the virtual contour field plane of the seawall site to obtain continuous wave height; S4: If any of the continuous wave heights exceeds the warning level, the seawall flooding warning alarm will be triggered.

[0026] Specifically, in step S1, the height information of different locations of the seawall should be measured at the seawall site. For specific measurement methods, please refer to Figure 2 , including the following steps: Sa1: Use a height measuring rod device to measure the fixed point height of the seawall, and record the physical length unit reading when the height measuring rod is used to measure the fixed point height; Sa2: The device captures images of the height measuring rod when measuring height, and records the image pixel unit readings when the height measuring rod device measures height at a fixed point; Sa3: Establish the corresponding relationship between the physical length unit reading and the image pixel unit reading when the height measuring rod device is used for fixed-point height measurement, and complete the process of measuring the height information of the seawall at a fixed point; Sa4: Repeat steps Sa1 to Sa3, perform the process of measuring the height information of the seawall at fixed points for no less than 3 times, and establish a corresponding relationship between the physical length unit reading and the image pixel unit reading when measuring the height at fixed points using no less than 3 height measuring rod devices.

[0027] Specifically, the wave semantic segmentation model in step S2 should call the specific modules in sequence to realize the function of the wave semantic segmentation model. For specific modules and calling sequence, please refer to Figure 5 , including the following modules: Data collection module 510: collects seawall overtopping image data for training and evaluation, and labels the collected seawall overtopping images; Model training module 520: used to train the semantic segmentation model so that it can accurately identify the wave coverage area in the seawall overtopping image; Model recognition module 530: used to perform semantic segmentation prediction on the collected seawall overtopping images and identify the wave coverage area.

[0028] Specifically, the data acquisition module 510 collects seawall overtopping image sets under different weather and sea conditions. And annotated dataset : The image collection of seawall overtopping And annotated dataset The collection process covers various sea conditions and lighting conditions. For specific considerations, please refer to Figure 5 , including the following considerations: Factor 1: Collection frequency: The collection frequency is set to once an hour to ensure timely monitoring and early warning of the seawall overtopping status; Factor 2: Weather factors: When collecting data, the impact of different weather conditions is taken into account, including sunny, cloudy, rainy, etc., to ensure that data collection is all-weather to meet the needs of seawall overtopping monitoring under different weather conditions; Factor 3: Day and night factors: When collecting data, the changes of day and night should be taken into account. The camera should have the function of night shooting. At night, additional lighting devices should be used to ensure image quality to ensure all-weather monitoring of seawall overtopping; Factor 4: Data desensitization and privacy processing: During the data collection process, data desensitization should be performed for parts involving personal privacy to ensure privacy security. The collection, storage and transmission of all data must comply with relevant privacy protection laws and standards; Factor 5: Data diversity: Data collection should cover different locations and angles of the seawall to ensure the robustness and generalization ability of the model. In addition, images under different lighting conditions should be collected to improve the adaptability of the model; Factor 6: Marking quality control: When marking the wave coverage area, take measures such as multi-person marking and marking result review to ensure the accuracy and consistency of the marking; Factor 7: Sea condition monitoring: During the data collection process, sea conditions, including wind speed, wind direction, tide and other information, are recorded for future analysis and model improvement; Factor 8: Data storage: The collected seawall overtopping image data is stored in a secure and reliable database for subsequent use in training and validating models; Factor 9: Data collection site setting: Select representative seawall overtopping monitoring sites for data collection, covering seawall overtopping conditions under different geographical conditions to ensure the universality and reliability of the model; The configuration details of each factor of the data acquisition module are shown in Table 1: Table 1 Configuration table of various factors of data acquisition module

[0029] Specifically, the model training module 520, for specific steps, please refer to Figure 5 , including the following steps: Step 521: Data preprocessing: And annotated dataset Preprocessing is performed, and the preprocessing methods include: image grayscale, image rotation, image reduction, image enlargement, image cropping, image displacement, histogram equalization, and high-pass filter kernel smoothing; Step 522: Model initialization: The CNN parameters of each level of the convolutional neural network of the wave semantic segmentation model are initialized using the Xaiver parameter initialization algorithm. The Xaiver parameter initialization algorithm is mathematically expressed as: (1) Where: Indicates from a arrive b Uniform distribution of is the number of input connections of the previous layer of the convolutional neural network; is the number of output connections of the next layer of the convolutional neural network.

[0030] Step 523: Model training: Using the pre-processed seawall overtopping image set And annotated dataset , the spray semantic segmentation model after model initialization is trained, and the model training loss function is the cross entropy loss function, which is mathematically expressed as: (2) Where: N The seawall overtopping image set And annotated dataset The number of samples; is the labeled dataset The actual annotations corresponding to the samples in The seawall overtopping image set The output obtained after the image corresponding to the sample in is input into the Langhua semantic segmentation model.

[0031] In this embodiment, the training curve diagram of the model training process 523 of the model training module 520 is shown in FIG. Fig.11 .

[0032] It can be understood by those skilled in the art that in this embodiment, the model obtained by training the model training process 523 of the model training module 520 uses the three indicators of accuracy, recall, and intersection-over-union to conduct a detailed evaluation of the recognition performance of the model in the wave-covered area and the non-wave-covered area. The evaluation data set used is the above-mentioned seawall overtopping image set. And annotated dataset .

[0033] The recognition performance evaluation results of the model trained in Example 1 are shown in Table 2: Table 2 Evaluation table of recognition performance of the model trained in Example 1

[0034] Specifically, the model recognition module 530, for specific steps, please refer to Figure 5 , including the following steps: Step 531: Data preprocessing: performing image preprocessing on the collected seawall overtopping images, including image grayscale processing, histogram equalization, and high-pass filter kernel smoothing; Step 532: Model recognition: Use the pre-processed seawall overtopping image to input the wave semantic segmentation model to obtain the wave coverage mask , in the waves covered with masks Any pixel in m Its mathematical expression is: (3) Where: It is the wave-covered mask; m It's a wave-covered mask A pixel in .

[0035] Specifically, the edge extraction algorithm in step S3 includes the following specific steps: S31: Initialization: Start from the upper left corner edge point of the image and traverse until the wave cover mask is found The first one to satisfy m= The pixel point of 1 is used as the starting point for edge extraction; S32: Record the current pixel: mark the current pixel as visited and record the coordinates of the current pixel .

[0036] S33: Go to the next pixel: Check the eight neighboring pixels above, below, left, right, and four diagonals of the current pixel in a clockwise or counterclockwise order, and select one neighboring pixel that has not been visited and satisfies m= The pixel with value 1 is taken as the next edge extraction point and marked as visited; S34: Repeat step S33: Circularly execute step S33 until returning to the starting point or finding no accessible pixels. If there are multiple unvisited pixels in the pixel neighborhood that satisfy m= 1, select the pixel farthest from the current pixel to ensure that the contour is traced as long as possible.

[0037] S35: Termination condition: The algorithm terminates after returning to the starting point, or when there are no more unvisited nodes and the m= The pixel is 1 and ends.

[0038] Specifically, the mathematical expression of the contour field coordinate mapping algorithm in step S3 is: (4) Where: is the coordinate of the wave edge extracted by the edge extraction algorithm in the image pixel unit; is the coordinate of the wave edge extracted by the edge extraction algorithm in physical length units; and The virtual contour plane of the seawall site is x Axis and y The result of isovalue field interpolation on the axis.

[0039] Specifically, the seawall flooding pre-alarm is triggered in step S4, and the pre-alarm mode is: when the wave height of any point in the continuous wave height obtained by visual measurement is greater than h Greater than the warning height , then the seawall flooding warning alarm is triggered, and its mathematical expression is: (5) Where: H is the continuous wave height obtained by visual measurement; h is the wave height at one point in the continuous wave height; is the pre-set alert altitude, and The value of is within [0m, 20m]. In this embodiment, The value is set to 4.5m.

[0040] According to whether there is overtopping of the seawall and whether there is overtopping exceeding the warning level, the seawall overtopping can be divided into three situations. The detailed classification is shown in Table 3: Table 3: Subdivision of three types of seawall overtopping

[0041] A video example with a duration of 30 seconds and a frame rate of 25 frames per second is processed using the method and configuration described in this embodiment, and the comprehensive judgment accuracy of the three situations of no overtopping and no warning, with overtopping and no warning, and with overtopping and warning is above 98%.

[0042] The accuracy of judging the three situations of seawall overtopping in Example 1 is shown in Table 4: Table 4. The accuracy of judging the three situations of seawall overtopping Example 2

[0043] Example 2 focuses on the assembly method of the device provided by the present invention and the details of its implementation at a seawall site. Figure 3 , Figure 4 and Figure 6a , Figure 6b , Figure 6cAs shown, the present invention provides an integrated device for visual monitoring of seawall overtopping waves that is compatible with the method for visual measurement and early warning of seawall overtopping wave height. The integrated device for visual monitoring of seawall overtopping waves consists of a camera, a terminal device server, a wireless communication network card and other structural parts. Its function is to shoot seawall overtopping wave images through a camera, and the terminal device server receives, processes and stores image data, and uses a semantic segmentation model to identify the wave coverage area, calculate the wave height in real time, and thus warn of the over-warning wave height, so as to provide real-time evaluation and early warning functions for seawall safety.

[0044] Specifically, the specific components of the integrated device for visual monitoring of seawall overtopping can be referred to Figure 3 , including a camera lens 301, a camera 302, a metal pole 303, a seawall wave overrunning visual monitoring terminal equipment box 304, a first supporting connection structure 305 for connecting the camera 302 and the metal pole 303 and supporting them, a second supporting connection structure 306 for connecting the seawall wave overrunning visual monitoring terminal equipment box 304 and the metal pole 303 and supporting them, and a pole base 307 for fixing the metal pole 303.

[0045] Specifically, the camera lens 301 is connected to the front of the camera 302, and together they constitute the visual perception equipment of the integrated device for visual monitoring of seawall overtopping waves, which has nighttime visual perception capabilities.

[0046] Specifically, the metal pole 303 plays a supporting and connecting role, and is connected to the camera 302 and the seawall overtopping visual monitoring terminal equipment box 304 through the first supporting connection structure 305 and the second supporting connection structure 306, and is fixed to the ground through the pole base 307.

[0047] Specifically, the specific structure of the seawall overtopping visual monitoring terminal equipment box 304 can be referred to Figure 4 It includes a terminal equipment box door 401, a terminal equipment box lock 402, a terminal equipment box body 403, a hinge 404 for connecting the terminal equipment box door 401 and the terminal equipment box body 403, a power supply socket 405, a terminal equipment network switch 406, a wireless communication network card 407, and a terminal equipment server 408. The seawall wave overrunning visual monitoring terminal equipment box 304 is connected to the metal pole 303 through a second supporting connection structure 306.

[0048] Specifically, the terminal device server 408 includes a processor and a memory; the memory stores at least one program, and the at least one program is loaded and executed by the processor to implement all or part of the steps in the various methods in Example 1.

[0049] Specifically, the terminal device server 408 includes a computer-readable storage medium; the computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for loading by a processor. A person skilled in the art can understand that all or part of the steps in the various methods in the above embodiment 1 can be completed by instructions, or by controlling related hardware through instructions, and the instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0050] In this embodiment, the integrated device for visual monitoring of seawall overtopping 604 is placed on the seawall platform 605, and its relative position relationship with respect to the seawall platform 607 to be observed is as follows: Figure 6a , Figure 6b , Figure 6c shown.

[0051] Specifically, the integrated device for visual monitoring of seawall overtopping waves 604 is placed on the seawall platform 605, and the seawall platform 607 to be observed is located in front of the camera in the positive direction of the integrated device for visual monitoring of seawall overtopping waves 604. In this embodiment, the straight-line distance between the integrated device for visual monitoring of seawall overtopping waves 604 and the seawall platform 607 to be observed is about 120m, and the sea surface 606 is between the two. Within this distance range, the camera's visual range 603 basically covers the seawall fixed-point measurement height point 601 and the seawall on-site virtual contour field plane 602. For details, please refer to Figure 6c .

[0052] Specifically, in this embodiment, there are 47 fixed-point measurement height points 601 for the seawall, which are evenly distributed in the virtual contour field plane 602 on the seawall site. The physical length unit heights corresponding to the 47 fixed-point measurement height points 601 for the seawall are measured by the measurement method for the fixed-point measurement of the height information of different positions of the seawall described in Example 1, and are associated with the image taken by the integrated device 604 for visual monitoring of seawall overtopping, to establish a corresponding relationship between the physical length unit readings and the image pixel unit readings during fixed-point height measurement by 47 pairs of height measuring rod devices.

[0053] The corresponding relationship between the physical length unit reading obtained by fixed-point height measurement and the image pixel unit reading in this embodiment is shown in Table 5: Table 5 Corresponding table of fixed-point height measurement of height measuring rod device

[0054] Specifically, by using the method of generating a virtual contour field plane of the seawall site using the contour field interpolation algorithm described in Example 1, and the corresponding relationship between the physical length unit reading and the image pixel unit reading when the 47 pairs of height measuring rods are used for fixed-point height measurement, a virtual contour field plane 602 of the seawall site is generated. In this embodiment, referring to Figure 6cThe horizontal width of the virtual contour field plane 602 of the seawall site is 90m, and the vertical height is 18m. Fig. 9 .

[0055] Specifically, in this embodiment, the implementation example device of the integrated device for visual monitoring of seawall overtopping 604 corresponds to Figure 7 , the implementation example device of the seawall overtopping visual monitoring terminal equipment box 304 corresponds to Figure 8 .

[0056] The above is a description of the embodiments of the present invention through specific implementations, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific implementations, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention.

Claims

1. A method for visual measurement and early warning of seawall overtopping height, characterized by: The following steps are involved: S1: Measure the height information of different positions of the seawall at the seawall site, and generate a virtual contour field plane of the seawall site by combining the contour field interpolation algorithm; S2: Visually monitor the seawall and use the wave semantic segmentation model to perform semantic segmentation and recognition on the images of seawall overtopping waves to obtain the wave coverage area; S3: Combining the virtual contour field plane of the seawall site and the semantic segmentation wave coverage, the edge extraction algorithm is used to extract the wave edge corresponding to the semantic segmentation wave coverage, and then the contour field coordinate mapping algorithm is used to map the wave edge coordinates to the virtual contour field plane of the seawall site to obtain continuous wave height; S4: If any of the continuous wave heights exceeds the warning level, the seawall flooding warning alarm will be triggered.

2. The method for visual measurement and early warning of seawall overtopping height according to claim 1 is characterized by: The method for measuring the height information of different positions of the seawall at a fixed point on the seawall site in step S1, the specific measurement method includes the following steps: Sa1: Use a height measuring rod device to measure the fixed point height of the seawall, and record the physical length unit reading when the height measuring rod is used to measure the fixed point height; Sa2: Take the image when the height measuring rod is used to measure the height, and record the image pixel unit reading when the height measuring rod device is used to measure the height at a fixed point; Sa3: Establish the corresponding relationship between the physical length unit reading and the image pixel unit reading when the height measuring rod device is used for fixed-point height measurement, and complete the process of measuring the height information of the seawall at a fixed point; Sa4: Repeat steps Sa1 to Sa3, perform the process of measuring the height information of the seawall at least 3 times, and establish the corresponding relationship between the physical length unit reading and the image pixel unit reading when measuring the height at a fixed point using at least 3 height measuring rod devices.

3. The method for visual measurement and early warning of seawall overtopping height according to claim 1 is characterized by: The wave semantic segmentation model in step S2 includes the following modules: Data collection module: collects seawall overtopping image data for training and evaluation, and annotates the collected seawall overtopping images; Model training module: used to train the semantic segmentation model so that it can accurately identify the wave-covered area in the seawall overtopping image; Model recognition module: used to perform semantic segmentation prediction on the collected seawall overtopping images and identify the wave coverage area.

4. The method for visual measurement and early warning of seawall overtopping height according to claim 3 is characterized by: The data acquisition module collects seawall overtopping image sets under different weather and sea conditions. And annotated dataset ; The seawall overtopping image set And annotated dataset The acquisition process covers various sea conditions and light conditions, as follows: Factor 1: Collection frequency: The collection frequency is set to once an hour to ensure timely monitoring and early warning of the seawall overtopping status; Factor 2: Weather factors: When collecting data, the impact of different weather conditions is taken into account, including sunny, cloudy, and rainy days, to ensure that data collection is all-weather to meet the needs of seawall overtopping monitoring under different weather conditions; Factor 3: Day and night factors: When collecting data, the changes of day and night should be taken into account. The camera should have the function of night shooting. At night, additional lighting devices should be used to ensure image quality to ensure all-weather monitoring of seawall overtopping; Factor 4: Data desensitization and privacy processing: During the data collection process, data desensitization should be performed for parts involving personal privacy to ensure privacy security. The collection, storage and transmission of all data must comply with relevant privacy protection laws and standards; Factor 5: Data diversity: Data collection should cover different locations and angles of the seawall to ensure the robustness and generalization ability of the model; in addition, images under different lighting conditions should be collected to improve the adaptability of the model; Factor 6: Marking quality control: When marking the wave coverage area, multiple people are required to mark and the marking results are reviewed to ensure the accuracy and consistency of the marking; Factor 7: Sea condition monitoring: During data collection, sea conditions, including wind speed, wind direction, and tide information, are recorded for future analysis and model improvement; Factor 8: Data storage: The collected seawall overtopping image data is stored in a secure and reliable database for subsequent use in training and validating models; Factor 9: Collection site setting: Select representative seawall overtopping monitoring sites for data collection, covering seawall overtopping conditions under different geographical conditions to ensure the universality and reliability of the model.

5. The method for visual measurement and early warning of seawall overtopping height according to claim 3 is characterized by: The implementation process of the model training module includes the following specific steps: S521: Data preprocessing: The seawall overtopping image set And annotated dataset Preprocessing is performed, and the preprocessing methods include: image grayscale, image rotation, image reduction, image enlargement, image cropping, image displacement, histogram equalization, and high-pass filter kernel smoothing; S522: Model initialization: The CNN parameters of each level of the convolutional neural network of the wave semantic segmentation model are initialized using the Xaiver parameter initialization algorithm. The Xaiver parameter initialization algorithm is mathematically expressed as: (1) Where: Indicates from a arrive b Uniform distribution of is the number of input connections of the previous layer of the convolutional neural network; is the number of output connections of the last layer of the convolutional neural network; S523: Model training: Using pre-processed seawall overtopping image set And annotated dataset , the spray semantic segmentation model after model initialization is trained, and the model training loss function is the cross entropy loss function, which is mathematically expressed as: (2) Where: N The seawall overtopping image set And annotated dataset The number of samples; is the labeled dataset The actual annotations corresponding to the samples in The seawall overtopping image set The output obtained after the image corresponding to the sample in is input into the Langhua semantic segmentation model.

6. The method for visual measurement and early warning of seawall overtopping height according to claim 3 is characterized by: The implementation process of the model identification module includes the following specific steps: S531: Data preprocessing: image preprocessing is performed on the collected seawall overtopping images, including image grayscale processing, histogram equalization, and high-pass filter kernel smoothing; S532: Model recognition: Use the pre-processed seawall overtopping image to input the wave semantic segmentation model to obtain the wave coverage mask , in the waves covered with masks Any pixel in m Its mathematical expression is: (3) Where: It is the wave-covered mask; m It's a wave-covered mask A pixel in .

7. The method for visual measurement and early warning of seawall overtopping height according to claim 1 is characterized by: The edge extraction algorithm in step S3 includes the following specific steps: S31: Initialization: Start from the upper left corner edge point of the image and traverse until the wave cover mask is found The first one to satisfy m= The pixel point of 1 is used as the starting point for edge extraction; S32: Record the current pixel: mark the current pixel as visited and record the coordinates of the current pixel ; S33: Go to the next pixel: Check the eight neighboring pixels in the up, down, left, right, and four diagonal directions of the current pixel in a clockwise or counterclockwise order, and select one neighboring pixel that has not been visited and satisfies m= The pixel with value 1 is taken as the next edge extraction point and marked as visited; S34: repeating step S33: looping through step S33 until returning to the starting point or finding no accessible pixels; If there are multiple unvisited pixel neighbors that satisfy m= 1, select the pixel farthest from the current pixel to ensure that the contour is traced as long as possible; S35: Termination condition: The algorithm terminates after returning to the starting point, or when there are no more unvisited nodes and the m= The pixel is 1 and ends.

8. The method for visual measurement and early warning of seawall overtopping height according to claim 1 is characterized by: The mathematical expression of the contour field coordinate mapping algorithm in step S3 is: (4) Where: is the coordinate of the wave edge extracted by the edge extraction algorithm in the image pixel unit; is the coordinate of the wave edge extracted by the edge extraction algorithm in physical length units; and The virtual contour plane of the seawall site is x Axis and y The result of isovalue field interpolation on the axis.

9. The method for visual measurement and early warning of seawall overtopping height according to claim 1, characterized in that: The triggering of the seawall flooding pre-alarm in step S4 is specifically carried out in the following manner: when the wave height at any point in the continuous wave heights obtained by visual measurement is h Greater than the warning height , then the seawall flooding warning alarm is triggered, and its mathematical expression is: (5) Where: H is the continuous wave height obtained by visual measurement; h is the wave height at one point in the continuous wave height; is the pre-set alert altitude, and The value of is in [0m, 30m].

10. An integrated device for visual monitoring of seawall overtopping using the method for visual measurement and early warning of seawall overtopping height according to any one of claims 1 to 9, characterized in that: It includes a camera lens, a camera, a metal pole, a terminal equipment box for visual monitoring of seawall overtopping waves, a supporting connection structure for connecting the camera and the metal pole and supporting the terminal equipment box for visual monitoring of seawall overtopping waves and the metal pole, and a pole base for fixing the metal pole; The terminal equipment box for visual monitoring of seawall overtopping waves includes a terminal equipment box door, a terminal equipment box lock, a terminal equipment box body, a hinge for connecting the terminal equipment box door and the terminal equipment box body, a power supply socket, a terminal equipment network switch, a wireless communication network card, and a terminal equipment server. The terminal equipment box for visual monitoring of seawall overtopping waves is connected to a metal pole through a supporting connection structure; The terminal device server includes a processor and a memory; the memory stores at least one program, and the at least one program is loaded and executed by the processor to implement the method for visual measurement and early warning of seawall overtopping height; The terminal device server includes a computer-readable storage medium; the computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for a processor to load to execute the method for visual measurement and early warning of seawall overtopping height.

Images