A method and system for identifying piping

Abstract

The present invention discloses a piping identification method and system, and relates to the technical field of water body identification. The method comprises: receiving infrared images sent back by a drone and generating a temperature matrix, automatically identifying and segmenting the boundary between the water body and shore plants based on a water body segmentation algorithm of a deep learning model; extracting temperature data of the water body area from the temperature matrix using the result of the water body segmentation; using an automated temperature identification algorithm, analyzing the processed temperature matrix, comparing the temperature differences between different areas inside the water body, and identifying the piping area; the present invention combines drone technology and deep learning methods to address the identification problem of piping under the water body, solves the challenge of accurately identifying and eliminating interference factors in a complex environment, uses a convolutional neural network to accurately segment the water body and shore plants, effectively eliminates the identification error caused by environmental interference, and makes the temperature data extraction more accurate, thereby improving the overall accuracy of piping identification.

Description

Technical Field

[0001] The present invention relates to the technical field of water body identification, and in particular to a piping identification method and system. Background Art

[0002] At present, there are indeed some technical bottlenecks in the identification of piping under water bodies. The biggest challenge is how to accurately identify and eliminate interference factors in a dynamic and complex environment. Existing technical methods mainly rely on drones equipped with infrared modules to take pictures of the target water body and combine temperature difference recognition algorithms to determine the piping situation. However, the accuracy of this method still needs to be improved when dealing with plants or environmental interference next to the water body. Specifically, due to the diversity and complexity of the natural environment, plants, buildings or other objects next to the water body will interfere with the infrared image, resulting in errors in the recognition results. Summary of the invention

[0003] The purpose of the present invention is to provide a piping identification method and system, which combines deep learning and image processing methods to achieve real-time identification of piping in water bodies, and effectively filter out interference information in a dynamic environment to achieve accurate monitoring.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] The present application provides a piping identification method, comprising the following steps:

[0006] Design route planning and create piping identification tasks, and automatically control the drone equipped with an infrared module to take photos of the target water body;

[0007] Receive the infrared images sent back by the drone and generate a temperature matrix. Based on the water segmentation algorithm of the deep learning model, automatically identify and segment the boundary between the water body and the shore plants.

[0008] Specifically, before receiving the infrared image sent back by the drone and generating the temperature matrix, it also includes:

[0009] Preprocessing of infrared image data, including normalization of the temperature matrix and conversion into time series format for time series analysis;

[0010] The standardization of the temperature matrix is ​​expressed as: ,in Expressed as a normalized temperature matrix, represents the temperature matrix, is expressed as the mean of the temperature matrix, is the standard deviation of the temperature matrix, represents the pixel coordinates in the temperature matrix;

[0011] The long short-term memory network (LSTM) model was trained using historical temperature data and optimized using mean square error as the loss function;

[0012] The loss function is expressed as: ,in is the actual temperature value, is the model prediction value, n is the length of the time series;

[0013] The trained LSTM model is used to predict the current temperature value and calculate the prediction error. When the prediction error exceeds the set threshold, the temperature anomaly in the piping will be identified and the piping risk warning will be triggered.

[0014] Using the results of water body segmentation, the temperature data of the water body area is extracted from the temperature matrix;

[0015] Using an automated temperature recognition algorithm, the processed temperature matrix is ​​analyzed and the piping area is identified by comparing the temperature differences between different areas within the water body.

[0016] Furthermore, the route planning is designed and the piping identification task is created, and the target water body is photographed by automatically controlling the drone equipped with an infrared module. The specific contents include:

[0017] The gray wolf optimization algorithm is used for route planning to determine the size of the gray wolf population and initialize the position of each gray wolf, where each gray wolf's position represents a different route planning scheme, and the position of the gray wolf is defined by the parameters of the route;

[0018] The drone flies along the planned route, uses an infrared camera to capture thermal image data of the water body, and transmits the thermal image data back to the ground station in real time through wireless communication technology.

[0019] Furthermore, the water segmentation algorithm based on the deep learning model automatically identifies and segments the boundary between the water body and the shore plants. The specific contents include:

[0020] Collect and annotate image datasets of water bodies and shore plants, and establish a CNN model, wherein the CNN model includes a convolutional layer, a pooling layer, and a fully connected layer;

[0021] The convolution layer is used to extract image features, the pooling layer is used to reduce the spatial dimension of the features, and the fully connected layer is used to segment the boundary between the water body and the shore plants;

[0022] Use the labeled data set to train the CNN model, and use the maximum inter-class variance to automatically determine the optimal threshold for image segmentation; segment the water body and shore plants;

[0023] The specific maximization of the inter-class variance is expressed as:

[0024] ;

[0025] in, and are the average grayscale values ​​of the pixels on both sides of the threshold T, and are the sum of the probabilities of the pixels on both sides of the threshold T, Represents grayscale value.

[0026] Furthermore, after automatically determining the optimal threshold for image segmentation by maximizing the inter-class variance, the method also includes: removing noise points and filling small holes in the water area by erosion and dilation operations;

[0027] The erosion operation is used to remove small noise points, and the pixel value after erosion is expressed as:

[0028] ;

[0029] in, is the pixel value of the original image, is a structural element, and Represents the coordinate drift of the structural element;

[0030] The dilation operation is used to fill small holes, and the pixel value after dilation is expressed as:

[0031] .

[0032] Furthermore, the temperature data of the water area is extracted from the temperature matrix, including:

[0033] The temperature data of the water area is extracted by performing element-by-element multiplication of the temperature matrix through the segmentation mask obtained by the deep learning model and maximizing the inter-class variance;

[0034] The element-by-element multiplication is expressed as: ;in Represents the filtered temperature data matrix, and the temperature data matrix of the water area is initially obtained. represents element-wise multiplication, is the original temperature matrix, is the segmentation mask matrix, where water areas are marked as 1 and non-water areas are marked as 0.

[0035] Furthermore, after extracting the temperature data of the water area from the temperature matrix, the temperature values ​​marked as non-water areas in the segmentation mask are cleared in the temperature matrix through conditional judgment, the temperature values ​​of the water area are retained, and the temperature interference of the non-water area is filtered out.

[0036] According to the temperature matrix after preliminary filtering Each element in Execution condition judgment:

[0037] When the mask matrix If the value is 1, then keep The value of

[0038] When the mask matrix If the value of is 0, The value of is cleared to zero;

[0039] Then store the result of conditional judgment in the final water area temperature data matrix middle.

[0040] Furthermore, the processed temperature matrix is ​​analyzed to identify the piping area by comparing the temperature differences in different areas of the water body. The specific contents include:

[0041] According to the filtered temperature matrix, the average temperature and temperature distribution of the water area are calculated;

[0042] The average temperature of the water area is expressed as: ;in, It's location The temperature value, is the area of ​​the water body, m and n are the number of rows and columns of the temperature matrix respectively;

[0043] The temperature distribution is represented by constructing a histogram, which is specifically represented as follows:

[0044] ;in It is a temperature barrel percentage, The temperature falls in the temperature bucket The number of pixels within the range;

[0045] According to the temperature distribution, a threshold is set to identify areas with abnormal temperature. For each pixel, the temperature difference with the surrounding pixels is calculated. When the threshold is exceeded, it is marked as a potential piping area.

[0046] Based on the identification results, a detailed identification report of the location and size of the piping is generated, and the identification report and warning information are pushed to the user end in real time.

[0047] Furthermore, based on the identification results, a detailed identification report of the piping location and size is generated, and the identification report and warning information are pushed to the user end in real time, including:

[0048] Based on the recognition results, potential piping areas are identified and a detailed recognition report is generated, which includes the specific location, size and related temperature anomaly data of the piping. Through real-time data communication technology, the recognition report and warning information will be immediately pushed to the user end, and a graphical interface will be used to display the location and distribution of the piping.

[0049] A piping identification system is applied to a piping identification method, comprising a route planning module, an image acquisition and preprocessing module, a water body segmentation module, a temperature analysis module, and a report generation and early warning module.

[0050] The route planning module uses the Gray Wolf Optimization Algorithm to design the route planning of the UAV, determine the altitude, speed and flight path of the route and create an inspection task, automatically control the UAV to take pictures of the target water body, and transmit the data back in real time;

[0051] The image acquisition and preprocessing module is used to receive the infrared image sent back by the drone and generate a temperature matrix, which also includes preprocessing of the image data;

[0052] The water body segmentation module, based on the water body segmentation algorithm of the deep learning model, automatically identifies and segments the boundary between the water body and the shore plants, determines the optimal threshold by maximizing the inter-class variance, and uses corrosion and expansion operations to remove noise and fill small holes in the water area;

[0053] The temperature analysis module uses the processed temperature matrix to calculate the average temperature and temperature distribution of the water area, and sets a threshold to identify the temperature abnormality area, and marks the potential piping area by comparing the temperature difference between different areas in the water body;

[0054] The report generation and warning module generates a detailed identification report of the piping location and size according to the identification result, and pushes the identification report and warning information to the user end through real-time data communication technology.

[0055] The beneficial effects of the present invention are:

[0056] The present invention combines drone technology and deep learning methods to solve the challenge of accurately identifying and eliminating interference factors in complex environments. Specifically, the drone is equipped with an infrared module to capture the thermal image data of the target water body in real time, and a convolutional neural network is used to accurately segment the water body and shore plants, effectively eliminating the recognition error caused by environmental interference, making the temperature data extraction more accurate, thereby improving the overall accuracy of piping identification.

[0057] By combining convolutional neural network (CNN) and gray wolf optimization algorithm, the present invention not only improves the recognition accuracy of water body boundaries, but also enhances the monitoring capability of pipe bursts. CNN can automatically process complex backgrounds, ensuring that even when the water body edge is not obvious or has irregular textures, it can still effectively distinguish between water and non-water areas. In addition, the use of gray wolf optimization algorithm for route planning enables the drone to quickly adapt to dynamic environments and obtain key data in real time. This dual strategy ensures real-time monitoring and identification of pipe bursts in complex environments.

[0058] Through real-time data feedback and intelligent analysis functions, the identification results are quickly pushed to the user end, which significantly improves the response speed to water safety management. The identification report includes the specific location, size and related temperature abnormality data of the pipe burst, ensuring that relevant personnel can obtain important information in a timely manner. The location and distribution of the pipe burst are displayed through a graphical interface, allowing users to intuitively identify risk areas. This real-time early warning mechanism not only improves the efficiency of pipe burst detection, but also enhances the accuracy of water safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0060] Figure 1 A schematic flow chart of a piping identification method provided in Example 1 of the present application;

[0061] Figure 2 A schematic diagram of a process for designing a route plan and creating a piping identification task for a piping identification method provided in Example 1 of the present application;

[0062] Figure 3 A schematic diagram of a flow chart of a piping identification method provided in Example 1 of the present application for automatically identifying and segmenting the boundary between a water body and shore plants;

[0063] Figure 4 A schematic diagram of a flow chart of identifying a piping area in a piping identification method provided in Example 1 of the present application;

[0064] Figure 5 A schematic diagram of the structure of a piping identification system provided in Example 2 of the present application. DETAILED DESCRIPTION

[0065] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0066] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0067] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0068] Example 1

[0069] See also Figure 1-Figure 5 This embodiment provides a piping identification method and system, which combines deep learning and image processing methods to achieve real-time identification of piping in water bodies, and effectively filters out interference information in a dynamic environment to achieve accurate monitoring.

[0070] The present invention provides a piping identification method, comprising the following steps:

[0071] S1. Design route planning and create piping identification tasks, and automatically control the drone equipped with an infrared module to take photos of the target water body;

[0072] Furthermore, the route planning is designed and the piping identification task is created, and the target water body is photographed by automatically controlling the drone equipped with an infrared module. The specific contents include:

[0073] S11. Use the gray wolf optimization algorithm to plan the route, determine the size of the gray wolf population and initialize the position of each gray wolf, where the position of each gray wolf represents a different route planning scheme, and the position of the gray wolf is defined by parameters such as the altitude, speed, and flight path of the route;

[0074] S12. The drone flies along the planned route, uses an infrared camera to capture thermal image data of the water body, and transmits the thermal image data back to the ground station in real time through wireless communication technology.

[0075] Specifically, by designing route planning and using drones to perform automatic inspection tasks, we can ensure that the target water body is photographed in a timely manner. This real-time nature enables the monitoring of the progress of pipe bursts to be completed as soon as possible. The infrared module carried by the drone can capture the thermal image data of the water body in real time during the flight, and quickly transmit this data back to the ground station through wireless communication technology. This process ensures the immediacy of data transmission, allowing subsequent analysis to be carried out as soon as possible.

[0076] The route planning is designed and route planning is performed through the Gray Wolf Optimization (GWO) algorithm. The specific contents include: determining the size of the gray wolf population and initializing the position of each gray wolf. Each position represents a potential route planning scheme. The position of each gray wolf is defined by multiple parameters, including the altitude, speed, flight path, etc. of the route; defining a fitness function to evaluate the effectiveness of each route. This function considers the total flight time, energy consumption, obstacle avoidance ability, etc. of the route. The goal is to find the optimal route so that the fitness function value is minimized; optimizing the search strategy by simulating the social hierarchy and hunting behavior of gray wolves. In the algorithm, each gray wolf tracks the positions of the three leading wolves Alpha, Beta, and Delta, and updates its own position according to their positions. This process involves collaboration and information sharing among gray wolves; in each iteration, according to the rules of the gray wolf algorithm, the position gradient of the current wolf pack is calculated and iterative search is performed. After each round of update, the wolf pack members (i.e., the route plan) may change to approach the optimal solution; in the wolf pack, the Alpha wolf represents the global optimal solution, and the Beta and Delta wolves represent the local optimal solution. Ordinary wolves in the wolf pack update their own positions according to the positions of Alpha, Beta, and Delta wolves to achieve global optimization; when the preset number of iterations is met or the path change is small enough, the algorithm has converged, and the optimal or suboptimal route planning solution can be obtained.

[0077] Specifically, the advantage of using the Gray Wolf Optimization (GWO) algorithm for route planning lies in its uniqueness in simulating the social hierarchy and hunting behavior of gray wolves, which enables the GWO algorithm to show fast convergence speed and high solution accuracy when solving complex optimization problems. The GWO algorithm effectively balances global search and local search by simulating the roles of the leader wolf (Alpha), follower wolves (Beta, Delta) and ordinary wolves (Omega) in the gray wolf group, thereby finding the global optimal solution in the multi-peak function. In addition, the GWO algorithm has a simple structure, few parameters to adjust, is easy to implement, and has a strong global search capability. These characteristics enable the GWO algorithm to quickly adapt to environmental changes in route planning, optimize routes in real time, and improve the inspection efficiency and safety of drones.

[0078] S2, receiving the infrared images sent back by the drone and generating a temperature matrix, and automatically identifying and segmenting the boundary between the water body and the shore plants based on the water body segmentation algorithm of the deep learning model;

[0079] The temperature matrix is ​​a two-dimensional array, in which each element represents the temperature value of the corresponding pixel in the image. In order to extract the temperature data of the water area from the temperature matrix, the water segmentation algorithm of the deep learning model is applied, which can automatically identify and segment the boundary between the water body and the shore plants. It can also achieve accurate water body boundary recognition even when the edge of the water body is not obvious or the surface has irregular textures. Compared with other feature learning models, such as Canny edge detection, the deep learning model is more effective in identifying water body boundaries in complex backgrounds, especially when facing water bodies with uneven temperatures.

[0080] Furthermore, before receiving the infrared image sent back by the UAV and generating the temperature matrix, it also includes:

[0081] Preprocess the infrared image data, including normalizing the temperature matrix and converting it into a time series format for time series analysis;

[0082] Specifically, the standardization process of the temperature matrix is ​​expressed as: ,in Expressed as a normalized temperature matrix, represents the temperature matrix, is expressed as the mean of the temperature matrix, is the standard deviation of the temperature matrix, Represents the pixel coordinates in the temperature matrix.

[0083] The long short-term memory network (LSTM) model was trained using historical temperature data and optimized using mean square error as the loss function;

[0084] Specifically, the loss function is expressed as: ,in is the actual temperature value, is the model prediction value and n is the length of the time series.

[0085] In the real-time data stream, the trained LSTM model is used to predict the current temperature value and calculate the prediction error. When the prediction error exceeds the set threshold, the temperature anomaly in the pipe burst will be identified and the pipe burst risk warning will be triggered.

[0086] Specifically, the prediction error is expressed as: .

[0087] Specifically, through the combination of deep learning models and anomaly detection mechanisms, the accuracy and timeliness of pipe burst identification can be significantly improved, the efficient execution of automatic inspection tasks can be achieved, and water safety can be effectively guaranteed.

[0088] In the present invention, the deep learning model includes convolutional neural network (CNN), deep belief network (DBN), Transformer model, attention mechanism model, etc.

[0089] Among them, the advantages of convolutional neural network (CNN) in piping recognition are mainly reflected in its powerful image feature extraction ability, automatic feature learning ability, and invariance to changes such as image translation and rotation. Through multi-layer convolution and pooling operations, CNN can automatically learn the key features to distinguish piping and non-piping areas from images without manual intervention, which greatly improves the efficiency and accuracy of recognition; in addition, the CNN model has strong generalization ability and can adapt to different environments and conditions, and can maintain a high recognition rate even in the case of poor image quality or occlusion. The following will take the convolutional neural network (CNN) as an example to illustrate.

[0090] Furthermore, the water segmentation algorithm based on the deep learning model automatically identifies and segments the boundary between the water body and the shore plants. The specific contents include:

[0091] S21, collecting and annotating image datasets of water bodies and shore plants, and establishing a CNN model, wherein the CNN model includes a convolutional layer, a pooling layer, and a fully connected layer;

[0092] The convolution layer is used to extract image features, the pooling layer is used to reduce the spatial dimension of the features, and the fully connected layer is used to segment the boundary between the water body and the shore plants;

[0093] S22. Use the labeled data set to train the CNN model, and use the maximum inter-class variance to automatically determine the optimal threshold for image segmentation; segment the water body and shore plants to provide accurate area definition for subsequent temperature analysis;

[0094] The specific maximization of the inter-class variance is expressed as:

[0095] ;

[0096] in, and are the average grayscale values ​​of the pixels on both sides of the threshold T, and are the sum of the probabilities of the pixels on both sides of the threshold T, Represents grayscale value.

[0097] Specifically, the water body segmentation algorithm based on the deep learning model can effectively identify and segment the boundaries between water bodies and shore plants by combining advanced deep learning technology and image processing methods. First, a large number of water body and shore plant image data sets are collected and annotated to train the deep CNN model, which contains convolutional layers, pooling layers, and fully connected layers. The convolutional layer is responsible for extracting key features in the image, such as texture and shape information; the pooling layer is used to reduce the spatial dimension of these features, thereby reducing the computational complexity; the fully connected layer is responsible for the final pixel classification task, that is, determining whether each pixel belongs to the water body or the shore plant; after the model training is completed, the method of maximizing the inter-class variance is further used to automatically determine the optimal threshold for image segmentation. By calculating the inter-class variance under different thresholds, the threshold that can maximize the distinction between water bodies and shore plants is found, thereby achieving accurate image segmentation. Specifically, the calculation of maximizing the inter-class variance involves the sum of the average grayscale value and probability of the pixels on both sides of the threshold T. The calculation method of maximizing the inter-class variance can effectively separate the water body from the shore plants, providing accurate regional definition for subsequent temperature analysis. This not only improves the accuracy of segmentation, but also reduces the impact of environmental interference on piping identification, thereby achieving high efficiency and high accuracy in piping identification in complex environments; especially in complex environments, it ensures accurate identification of water body boundaries. The automation and efficiency of this process improves the ability of real-time detection.

[0098] Furthermore, after automatically determining the optimal threshold for image segmentation by maximizing the inter-class variance, the method also includes: removing noise points and filling small holes in the water area through corrosion and expansion operations; making the boundaries of the water area clearer and reducing outliers in the data;

[0099] The erosion operation is used to remove small noise points, and the pixel value after erosion is expressed as:

[0100] ;

[0101] in, is the pixel value of the original image, is a structural element, and Represents the coordinate drift of the structural element;

[0102] The erosion operation updates the center pixel value by taking the minimum of all pixel values ​​covered by the structure element. This helps eliminate small white areas (noise) and shrink larger white areas.

[0103] The dilation operation is used to fill small holes to make the boundary of the water area clearer. The pixel value after dilation is expressed as:

[0104] ; The dilation operation updates the center pixel value by taking the maximum value of all the pixel values ​​covered by the structure element. This helps connect adjacent white areas and fill small black areas (holes).

[0105] Specifically, the effect of the expansion operation is mainly reflected in enhancing the white area in the image and making it more coherent. By covering it with structural elements and taking the maximum pixel value, it can effectively fill the small holes and breaks in the image, making the boundary of the water area clearer and more continuous. This is especially important for subsequent image analysis, as it can provide more accurate regional definition for continuous and precise water boundaries, making image-based temperature analysis and other related calculations more reliable. In addition, the expansion operation can also help eliminate noise and small interferences, further improve the quality of the image, and lay a solid foundation for in-depth image analysis and feature extraction.

[0106] S3. Using the results of water body segmentation, the temperature data of the water body area is extracted from the temperature matrix; when the edge of the water body is not obvious or the surface has irregular texture, the temperature interference of the non-water body part is accurately filtered out.

[0107] Furthermore, the temperature data of the water area is extracted from the temperature matrix, including:

[0108] The temperature data of the water area is extracted by performing element-by-element multiplication of the temperature matrix through the segmentation mask obtained by the deep learning model and maximizing the inter-class variance;

[0109] Element-wise multiplication is expressed as: ;in Represents the filtered temperature data matrix, and the temperature data matrix of the water area is initially obtained. represents element-wise multiplication, is the original temperature matrix, is the segmentation mask matrix, where water areas are marked as 1 and non-water areas are marked as 0.

[0110] Specifically, the deep learning model and the segmentation mask obtained by maximizing the inter-class variance can effectively extract the temperature data of the water body area from the temperature matrix. The element-by-element multiplication operation combines the original temperature matrix with the segmentation mask so that only the temperature values ​​of the areas marked as water bodies are retained, and the temperature values ​​of the non-water body parts are cleared. The final effect is to obtain a filtered temperature data matrix that focuses on the temperature distribution of the water body area, providing accurate basic data for subsequent temperature analysis and piping identification, thereby enhancing the reliability and effectiveness of the analysis.

[0111] Furthermore, after extracting the temperature data of the water area from the temperature matrix, the temperature values ​​marked as non-water areas (i.e., the mask value is 0) in the segmentation mask are cleared in the temperature matrix through conditional judgment, and only the temperature values ​​of the water area (mask value is 1) are retained, thereby effectively filtering the temperature interference of the non-water area.

[0112] According to the temperature matrix after preliminary filtering Each element in Execution condition judgment:

[0113] When the mask matrix The value of is 1 (indicating that the location belongs to the water area), then keep The value of

[0114] When the mask matrix If the value of is 0 (indicating that the location is a non-water area), The value of is cleared to zero;

[0115] Then store the result of conditional judgment in the final water area temperature data matrix middle.

[0116] Specifically, the temperature interference of the non-water body part is effectively filtered out through conditional judgment; specifically, conditional judgment is performed on each element in the temperature matrix after preliminary filtering: when the value of the segmentation mask is 1 (indicating that the location belongs to the water body area), the corresponding temperature value is retained; when the mask value is 0 (indicating that the location belongs to the non-water body area), the corresponding temperature value is cleared. Finally, a matrix containing only the temperature values ​​of the water body area is generated, ensuring that the temperature interference of the non-water body part is effectively eliminated. This precise filtering not only improves the accuracy of the temperature data in the water body area, but also provides a reliable basis for the subsequent piping identification and analysis, and enhances the overall monitoring effect and reliability.

[0117] S4. Analyze the processed temperature matrix using an automated temperature recognition algorithm and identify piping areas by comparing the temperature differences between different areas within the water body;

[0118] Furthermore, the processed temperature matrix is ​​analyzed to identify the piping area by comparing the temperature differences in different areas of the water body. The specific contents include:

[0119] S41, calculating the average temperature and temperature distribution of the water area according to the filtered temperature matrix;

[0120] The average temperature of the water area is expressed as: ;in, It's location The temperature value, is the area of ​​the water body, m andn are the number of rows and columns of the temperature matrix, respectively.

[0121] The temperature distribution is represented by constructing a histogram, which is specifically represented as follows:

[0122] ;in It is a temperature barrel percentage, The temperature falls in the temperature bucket The number of pixels in the range.

[0123] S42. According to the temperature distribution, a threshold is set to identify the area with abnormal temperature. For each pixel, the temperature difference with the surrounding pixels is calculated. When the temperature exceeds the threshold, it is marked as a potential piping area.

[0124] Specifically, by analyzing the processed temperature matrix, the average temperature of the water area is calculated, which provides a benchmark for the overall temperature level. Then, by constructing a histogram to display the temperature distribution, the frequency of different temperature values ​​in the water body is understood, thereby identifying temperature anomaly areas. Specifically, by setting a threshold and calculating the temperature difference between each pixel and its surrounding pixels, when these differences exceed the set threshold, these areas are marked as potential piping areas. This method can accurately locate piping areas, improve the accuracy and timeliness of piping detection, and provide a scientific basis for taking preventive measures.

[0125] S5. Generate a detailed identification report of the location and size of the piping burst based on the identification results, and push the identification report and warning information to the user end in real time.

[0126] Furthermore, based on the identification results, a detailed identification report of the piping location and size is generated, and the identification report and warning information are pushed to the user end in real time, including:

[0127] According to the identification results, potential piping areas are identified and a detailed identification report is generated, which includes the specific location, size and related temperature anomaly data of piping. Through real-time data communication technology, the identification report and warning information will be immediately pushed to the user end to ensure that relevant personnel can obtain important information in a timely manner. Specifically, a graphical interface is used to display the location and distribution of piping, so that users can intuitively identify risk areas, thereby providing a basis for subsequent monitoring and emergency response. It not only improves the efficiency of piping detection, but also enhances the real-time and accuracy of water safety management.

[0128] Overall, the piping identification method of the present invention realizes real-time monitoring and identification of piping in water bodies. First, the route planning is designed using the Gray Wolf optimization algorithm, and the target water body is photographed by an infrared module mounted on a drone to ensure timely data transmission. Next, the infrared image transmitted back by the drone is received, a temperature matrix is ​​generated, and a water body segmentation algorithm based on a deep learning model is applied to automatically identify the boundary between the water body and the shore plants. The temperature interference of the non-water body part is filtered out by element-by-element multiplication operation, and the temperature data of the water body area is extracted. Subsequently, the temperature matrix is ​​analyzed using an automated temperature recognition algorithm, the average temperature and temperature distribution of the water body are calculated, and the temperature abnormality area is identified by setting a threshold, marking the potential piping area. Finally, a detailed identification report is generated, including the specific location and size of the piping, and the warning information is pushed to the user end in real time to ensure that relevant personnel can obtain and intuitively identify the risk area in a timely manner, thereby improving the efficiency of piping detection and the accuracy of water safety management.

[0129] Example 2

[0130] This embodiment provides a piping identification system, which is applied to a piping identification method, and the specific contents include:

[0131] The route planning module uses the Gray Wolf optimization algorithm to design the route planning of the drone, determine the altitude, speed and flight path of the route and create inspection tasks. By automatically controlling the drone, it ensures that the target water body can be photographed and the data can be transmitted back in real time.

[0132] The image acquisition and preprocessing module is used to receive infrared images sent back by the drone and generate a temperature matrix. In this process, it also includes preprocessing of image data, such as standardization and time series format conversion, to provide basic data for subsequent analysis;

[0133] The water segmentation module uses a water segmentation algorithm based on a deep learning model to automatically identify and segment the boundaries between water bodies and shore plants. It determines the optimal threshold by maximizing the inter-class variance, and uses corrosion and expansion operations to remove noise and fill small holes in the water area, ensuring that the boundaries of the water area are clearer.

[0134] The temperature analysis module uses the processed temperature matrix to calculate the average temperature and temperature distribution of the water area, and sets thresholds to identify temperature anomaly areas. By comparing the temperature differences between different areas in the water body, potential piping areas are marked to provide data support for subsequent risk assessment;

[0135] The report generation and warning module generates a detailed identification report of the location and size of the piping according to the identification results, and pushes the identification report and warning information to the user end through real-time data communication technology, ensuring that relevant personnel can obtain important information in a timely manner, thereby improving the efficiency of piping detection and the accuracy of water safety management.

[0136] This piping identification system realizes efficient monitoring and identification of piping in water bodies, provides a route planning module to ensure that drones can accurately inspect target water bodies, and the image acquisition and preprocessing module provides high-quality temperature data for subsequent analysis. The water body segmentation module uses deep learning technology to accurately identify the boundary between water bodies and shore plants, thereby improving segmentation accuracy; the temperature analysis module calculates temperature differences and timely marks potential piping areas, providing a scientific basis; finally, the report generation and warning module ensures that relevant personnel can obtain detailed identification reports and warning information in a timely manner, thereby significantly improving the efficiency of piping detection and the accuracy of water safety management, and ensuring the safety and stability of the water area.

[0137] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A piping identification method, characterized in that: The following steps are involved: Design route planning and create piping identification tasks, and automatically control the drone equipped with an infrared module to take photos of the target water body; Receive the infrared images sent back by the drone and generate a temperature matrix. Based on the water segmentation algorithm of the deep learning model, automatically identify and segment the boundary between the water body and the shore plants. Specifically, before receiving the infrared image sent back by the drone and generating the temperature matrix, it also includes: Preprocessing of infrared image data, including normalization of the temperature matrix and conversion into time series format for time series analysis; The temperature matrix is ​​standardized and expressed as: ,in Expressed as a normalized temperature matrix, represents the temperature matrix, is expressed as the mean of the temperature matrix, is the standard deviation of the temperature matrix, represents the pixel coordinates in the temperature matrix; Use historical temperature data to train the LSTM model and use mean square error as the loss function for optimization; The loss function is expressed as: ,in is the actual temperature value, is the model prediction value, n is the length of the time series; The trained LSTM model is used to predict the current temperature value and calculate the prediction error. When the prediction error exceeds the set threshold, the temperature anomaly in the piping will be identified and the piping risk warning will be triggered. Using the results of water body segmentation, the temperature data of the water body area is extracted from the temperature matrix; Utilize the automated temperature recognition algorithm to analyze the processed temperature matrix and identify the piping area by comparing the temperature differences in different areas within the water body; Among them, the temperature data of the water area is extracted from the temperature matrix, including: The temperature data of the water area is extracted by performing element-by-element multiplication of the temperature matrix through the segmentation mask obtained by the deep learning model and maximizing the inter-class variance; The element-by-element multiplication is expressed as: ;in Represents the filtered temperature data matrix, and the temperature data matrix of the water area is initially obtained. represents element-wise multiplication, is the original temperature matrix, is the segmentation mask matrix, where the water area is marked as 1 and the non-water area is marked as 0; The processed temperature matrix is ​​analyzed to identify piping areas by comparing the temperature differences in different areas of the water body, including: According to the filtered temperature matrix, the average temperature and temperature distribution of the water area are calculated; The average temperature of the water area It is expressed as: ;in, It's location The temperature value, is the area of ​​the water body, m and n points Specifically, the number of rows and columns of the temperature matrix; The temperature distribution is represented by constructing a histogram, which is specifically represented as follows: ;in It is a temperature barrel percentage, The temperature falls in the temperature bucket The number of pixels within the range; According to the temperature distribution, a threshold is set to identify areas with abnormal temperature. For each pixel, the temperature difference with the surrounding pixels is calculated. When the threshold is exceeded, it is marked as a potential piping area.

2. A piping identification method according to claim 1, characterized in that: The design of route planning and creation of piping identification tasks involves automatically controlling the drone equipped with an infrared module to take photos of the target water body. The specific contents include: The gray wolf optimization algorithm is used for route planning to determine the size of the gray wolf population and initialize the position of each gray wolf, where each gray wolf's position represents a different route planning scheme, and the position of the gray wolf is defined by the parameters of the route; The drone flies along the planned route, uses an infrared camera to capture thermal image data of the water body, and transmits the thermal image data back to the ground station in real time through wireless communication technology.

3. A piping identification method according to claim 1, characterized in that: The water segmentation algorithm based on the deep learning model automatically identifies and segments the boundary between the water body and the shore plants. The specific contents include: Collect and annotate image datasets of water bodies and shore plants, and establish a CNN model, wherein the CNN model includes a convolutional layer, a pooling layer, and a fully connected layer; The convolution layer is used to extract image features, the pooling layer is used to reduce the spatial dimension of the features, and the fully connected layer is used to segment the boundary between the water body and the shore plants; Use the labeled data set to train the CNN model, and use the maximum inter-class variance to automatically determine the optimal threshold for image segmentation; segment the water body and shore plants; Specifically maximize the between-class variance It is expressed as: ； in, and are the average grayscale values ​​of the pixels on both sides of the threshold T, and are the sum of the probabilities of the pixels on both sides of the threshold T, Represents grayscale value.

4. A piping identification method according to claim 3, characterized in that: After automatically determining the optimal threshold for image segmentation using maximization of inter-class variance, it also includes: removing noise and filling small holes in water areas through erosion and dilation operations; The erosion operation is used to remove small noise points. The pixel value after erosion It is expressed as: ； in, is the pixel value of the original image, is a structural element, and Represents the coordinate drift of the structural element; The dilation operation is used to fill small holes. The pixel value after dilation It is expressed as: 。 5. A piping identification method according to claim 4, characterized in that: After extracting the temperature data of the water area from the temperature matrix, the temperature values ​​marked as non-water areas in the segmentation mask are cleared in the temperature matrix through conditional judgment, the temperature values ​​of the water area are retained, and the temperature interference of the non-water area is filtered out. According to the temperature matrix after preliminary filtering Each element in Execution condition judgment: When the mask matrix If the value is 1, then keep The value of When the mask matrix If the value of is 0, The value of is cleared to zero; Then store the result of conditional judgment in the final water area temperature data matrix middle.

6. A piping identification method according to claim 1, characterized in that: Also includes: Based on the identification results, a detailed identification report of the location and size of the piping is generated, and the identification report and warning information are pushed to the user end in real time.

7. A piping identification method according to claim 6, characterized in that: Based on the identification results, a detailed identification report of the piping location and size is generated, and the identification report and warning information are pushed to the user end in real time, including: Based on the recognition results, potential piping areas are identified and a detailed recognition report is generated, which includes the specific location, size and related temperature anomaly data of the piping. The recognition report and warning information are pushed to the user end through real-time data communication technology. A graphical interface is used to display the location and distribution of the piping.

8. A piping identification system, using a piping identification method according to any one of claims 1 to 6, characterized in that: It includes route planning module, image acquisition and preprocessing module, water body segmentation module, temperature analysis module and report generation and early warning module. The route planning module uses the Gray Wolf Optimization Algorithm to design the route planning of the UAV, determine the altitude, speed and flight path of the route and create an inspection task, automatically control the UAV to take pictures of the target water body, and transmit the data back in real time; The image acquisition and preprocessing module is used to receive the infrared image sent back by the drone and generate a temperature matrix, which also includes preprocessing of the image data; The water body segmentation module, based on the water body segmentation algorithm of the deep learning model, automatically identifies and segments the boundary between the water body and the shore plants, determines the optimal threshold by maximizing the inter-class variance, and uses corrosion and expansion operations to remove noise and fill small holes in the water area; The temperature analysis module uses the processed temperature matrix to calculate the average temperature and temperature distribution of the water area, and sets a threshold to identify the temperature abnormality area, and marks the potential piping area by comparing the temperature difference between different areas in the water body; The report generation and warning module generates a detailed identification report of the piping location and size according to the identification result, and pushes the identification report and warning information to the user end through real-time data communication technology.

Images