Municipal water-filled main pipe internal defect detection method, equipment and medium

Through the optimization of feature optimization strategies and deep learning models, the problems of low efficiency and poor accuracy in traditional detection technology are solved, and efficient and accurate internal defect detection of municipal full water main pipes is achieved, adapting to the detection needs of complex drainage pipelines and ensuring urban drainage systems healthy operation.

CN118941506APending Publication Date: 2024-11-12POWERCHINA WATER ENVIRONMENT GOVERANCE +3
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410930517.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The internal defect detection technology of traditional municipal full water main pipe is low efficiency, poor accuracy, and has a large resource consumption. The deep learning model is limited in feature selection and data quality, resulting in insufficient robustness and generalization capabilities, making it difficult to adapt to complex and changeable drainage pipelines. Test requirements.

Method used

Feature optimization strategies, including time-frequency analysis algorithm, adaptive window function and search algorithm, convert one-dimensional signals into two-dimensional feature maps, combine reverse learning strategies and Levi flight and Brownian motion to optimize deep learning models and improve detection efficiency and precision.

Benefits of technology

It significantly improves the efficiency and accuracy of internal defect detection of full water main pipes, shortens the detection cycle, reduces the risk of human misjudgment, and provides strong technical support for the maintenance and management of municipal facilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118941506A_ABST
    Figure CN118941506A_ABST
Patent Text Reader

Abstract

The invention discloses a municipal full-water main pipe internal defect detection method and device and a medium, internal defect data of a full-water main pipe is obtained firstly, and then the internal defect data is represented by using a feature optimization strategy to obtain an internal defect feature map. And further, constructing an initial full-water main pipe internal defect detection model, and training the detection model based on the internal defect feature map to obtain the full-water main pipe internal defect detection model. A to-be-detected internal defect feature map is input into the trained full-water main pipe internal defect detection model for defect detection, and a defect detection result can be rapidly and accurately obtained. According to the embodiment of the invention, the problems of low efficiency, poor precision and the like of the traditional detection method in the detection of the internal defects of the full-water main pipe are solved, so that the efficiency and the precision of the detection of the internal defects of the full-water main pipe are effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of the present application relate to but are not limited to the field of equipment detection, and in particular to a method, equipment and medium for detecting internal defects of a municipal full water main. Background Art

[0002] Municipal drainage pipes are the key to urban infrastructure. Their operating environment is complex and changeable. In addition, the irregularities in urban construction and management often make it difficult to find and maintain internal defects in the pipes in a timely manner, leading to serious drainage problems. The causes of defects are complex, covering external factors such as burial depth, soil properties, and traffic load, as well as internal factors such as pipe material aging and corrosion, which are intertwined to aggravate drainage deterioration.

[0003] Traditional internal defect detection technology for full water mains relies on a lot of manpower, material resources and expert experience, which is inefficient and consumes a lot of resources. With the leap of data technology, defect detection models have emerged, which can automatically learn fault characteristics and reduce the reliance on prior knowledge, becoming a new favorite in research. However, practical applications require a balance between detection efficiency, effect and resource consumption, and traditional methods often lose sight of one while focusing on the other.

[0004] For internal defect detection methods based on feature extraction, the quality of feature extraction directly affects the subsequent results. However, some existing feature extraction methods often have good theoretical effects, but poor robustness in actual defect processing, and cannot achieve universality. In addition, the subjectivity of manual feature selection affects the detection stability and generalization ability. Although deep learning models are powerful, they are limited by the quality of early data and feature selection, and the resource consumption increases due to the increase in the number of layers, which faces challenges in practical applications.

[0005] Therefore, there is an urgent need to develop efficient, accurate and resource-friendly internal defect detection technology, combine intelligent feature extraction and deep learning optimization, ensure data quality, improve model generalization ability, and control resource consumption at the same time, in order to cope with the complex and changeable municipal drainage pipeline inspection needs and ensure the healthy operation of the urban drainage system. Summary of the invention

[0006] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0007] The embodiment of the present application provides a method for detecting internal defects of a municipal full water main, which can effectively improve the efficiency and accuracy of detecting internal defects of a full water main.

[0008] On the one hand, an embodiment of the present application provides a method for detecting internal defects of a municipal full water main, comprising:

[0009] Obtain internal defect data of municipal full water mains;

[0010] Characterizing the internal defect data of the municipal full water main using a feature optimization strategy to obtain an internal defect feature map;

[0011] Construct an internal defect detection model for an initially full water main;

[0012] Training the initial full water main internal defect detection model based on the internal defect feature map to obtain a trained full water main internal defect detection model;

[0013] Obtaining a characteristic diagram of internal defects to be detected;

[0014] The internal defect feature map to be detected is input into the trained full water main internal defect detection model to perform defect detection and obtain a detection result.

[0015] In one embodiment of the present application, the feature optimization strategy includes a time-frequency analysis algorithm, an adaptive window function, and a search algorithm. The feature optimization strategy is used to characterize the internal defect data of the municipal full water main to obtain an internal defect feature map, including: using the time-frequency analysis algorithm to perform time-frequency analysis and parameter extraction on the internal defect data to obtain a time-frequency aggregation degree; obtaining the standard deviation of the adaptive window function; adaptively matching the standard deviation of the adaptive window function according to the time-frequency aggregation degree and the search algorithm until the standard deviation reaches a preset matching threshold, thereby obtaining an internal defect feature map;

[0016] Among them, the expressions of adaptive window function, time-frequency analysis algorithm, search algorithm and time-frequency aggregation are:

[0017]

[0018] Among them, x(u) is the collected signal, u is the time variable, j is the imaginary unit of the complex number, σ t is the standard deviation, ω is the frequency, t is the time, and D is the window length.

[0019] In one embodiment of the present application, the time-frequency analysis and parameter extraction of the internal defect data using a time-frequency analysis algorithm to obtain a time-frequency aggregation degree includes: analyzing the internal defect data using a time-frequency analysis method to obtain a defect signal in the time-frequency domain; and extracting the time-frequency aggregation degree from the defect signal in the time-frequency domain.

[0020] In one embodiment of the present application, the internal defect feature map to be detected is input into the trained full water main internal defect detection model for defect detection to obtain a detection result, including: performing feature processing on the internal defect feature map to be detected to obtain an enhanced feature map; performing dimensionality reduction and transformation on the enhanced feature map to obtain a reduced dimensionality feature map; and mapping and normalizing the reduced dimensionality feature map to obtain a detection result.

[0021] In one embodiment of the present application, the feature processing of the internal defect feature map to be detected to obtain an enhanced feature map includes: using a convolution kernel for each channel of the internal defect feature map to be detected, splicing the outputs of the convolution kernel to obtain a feature map; and performing channel fusion on the feature map to obtain an enhanced feature map.

[0022] In one embodiment of the present application, the method further includes: when the detection result meets the preset result, using the detection result as the output result; when the detection result is judged to fail to meet the preset result, correcting the weight of the full water main internal defect detection model.

[0023] In one embodiment of the present application, the feature optimization strategy includes a reverse learning strategy, Levy flight and Brownian motion, and the mathematical expressions of the reverse learning strategy, Levy flight and Brownian motion are:

[0024]

[0025] X i =X i-1 +α⊕levy(β),

[0026]

[0027] X k =X k-1 +rand(),

[0028] Among them, a represents the previous level of optimized parameters, b represents the next level of optimized parameters, and X i represents the i-th individual in the population, α is the step factor, ⊕ represents the dot product, rand() represents the generation of a random number between 0 and 1, β=1.5, Γ is the standard Gamma function, and v obeys the standard normal distribution.

[0029] In one embodiment of the present application, the initial full water main internal defect detection model is trained based on the internal defect feature map to obtain a trained full water main internal defect detection model, including: constructing a training set and a validation set according to the internal defect feature map; using the training set and the validation set to respectively train and validate the initial full water main internal defect detection model to obtain the full water main internal defect detection model.

[0030] On the other hand, an embodiment of the present application further provides an electronic device, including:

[0031] at least one processor;

[0032] at least one memory for storing at least one program;

[0033] When at least one of the programs is executed by at least one of the processors, the information processing method of the collaborative whiteboard as described above is implemented.

[0034] On the other hand, an embodiment of the present application further provides a computer-readable storage medium, which stores a computer program executable by a processor. When the computer program executable by the processor is executed by the processor, it is used to implement the information processing method of the collaborative whiteboard as described above.

[0035] The embodiment of the present application provides a method for detecting internal defects of a municipal full water main pipe. The method first obtains the internal defect data of the municipal full water main pipe, and then uses the feature optimization strategy to characterize the internal defect data of the municipal full water main pipe to obtain an internal defect feature map. The one-dimensional signal (internal defect data of the full water main pipe) is converted into a two-dimensional image (internal defect feature map) by the feature optimization strategy. This process effectively extracts the key features of the defect, so that the subsequent detection model can more accurately identify and analyze the type of defect and other information, thereby improving the efficiency of data processing, and further enhancing the accuracy of the detection results. Further, an initial full water main pipe internal defect detection model is constructed, and the initial full water main pipe internal defect detection model is trained based on the internal defect feature map to obtain a full water main pipe internal defect detection model. By inputting the acquired internal defect feature map to be detected into the trained full water main pipe internal defect detection model for defect detection, the defect detection results can be obtained quickly and accurately, which greatly shortens the detection cycle, reduces the risk of human misjudgment, and provides strong technical support for the maintenance and management of municipal facilities. The embodiments of the present application solve the problems of low efficiency and poor accuracy faced by traditional detection methods in detecting internal defects of full water mains, thereby effectively improving the efficiency and accuracy of internal defect detection of full water mains. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of a method for detecting internal defects of a municipal full water main provided by an embodiment of the present application;

[0037] Figure 2 is a schematic diagram of specific steps of step 120 provided in an embodiment of the present application;

[0038] Figure 3 is a schematic diagram of specific steps of step 160 provided in an embodiment of the present application;

[0039] Figure 4 It is an optimized Densenet network structure diagram provided in an embodiment of the present application;

[0040] Figure 5 It is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0042] It should be noted that, although the logical order is shown in the flowchart, in some cases, the steps shown or described can be performed in a different order from that in the flowchart. The terms "first", "second", etc. in the specification, claims and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the limiting conditions that the present invention can be implemented, so they have no technical substantive significance. Any modification of the structure, change of the proportional relationship or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope of the technical content disclosed by the present invention. At the same time, the terms such as "upper", "lower", "left", "right", "middle" and "one" quoted in this specification are only for the convenience of narration, and are not used to limit the scope of the implementation of the present invention. The change or adjustment of the relative relationship should also be regarded as the scope of the implementation of the present invention without substantially changing the technical content.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0044] Traditional internal defect detection technology for full water mains relies on a lot of manpower, material resources and expert experience, which is inefficient and consumes a lot of resources. With the leap of data technology, defect detection models have emerged, which can automatically learn fault features and reduce the dependence on prior knowledge, becoming a new favorite in research. However, practical applications need to balance detection efficiency, effect and resource consumption, and traditional methods often lose one while focusing on the other. For internal defect detection methods based on feature extraction, the quality of feature extraction directly affects the subsequent effect. However, some existing feature extraction methods often have good theoretical effects, but poor robustness in actual defect processing, and cannot be universal, and the subjectivity of manual feature selection affects the detection stability and generalization ability. Although deep learning models are powerful, they are limited by the quality of early data and feature selection, and the increase in the number of layers leads to a surge in resource consumption, which faces challenges in practical applications. Therefore, it is urgent to develop efficient, accurate and resource-friendly internal defect detection technology, combining intelligent feature extraction and deep learning optimization to ensure data quality, improve model generalization ability, and control resource consumption at the same time, in order to cope with the complex and changeable municipal drainage pipeline detection needs and ensure the healthy operation of urban drainage systems.

[0045] In view of this, the embodiment of the present application provides a method for detecting internal defects of a municipal full water main, an electronic device and a computer-readable storage medium. The method first collects defect data inside the municipal full water main, which are presented in the form of one-dimensional signals. Subsequently, the feature optimization strategy is used to convert these one-dimensional signals into intuitive and information-rich two-dimensional images, namely, internal defect feature maps. This conversion process not only significantly extracts the key features of the defects, but also greatly enhances the readability of the data, providing a more accurate input basis for subsequent detection models, thereby achieving faster and more accurate identification and analysis of defect types and characteristics, and effectively improving the overall efficiency of data processing and the accuracy of detection results. Next, an initial full water main internal defect detection model is constructed, and the model is trained based on the internal defect feature map. This process ensures that the model can fully learn the complex features of the defects and obtain an efficient and accurate full water main internal defect model. The internal defect feature map to be detected is input into the trained full water main internal defect model, and the model can respond quickly and accurately output defect detection results. This process greatly shortens the detection cycle, reduces the risk of misjudgment due to human factors, and provides strong technical support and guarantee for the maintenance and management of municipal facilities. The embodiment of the present application solves the key problems of low efficiency and insufficient precision in the traditional internal defect detection methods of full water mains, achieves a double leap in detection efficiency and precision, and opens up a new path for the safe operation and efficient management of municipal facilities.

[0046] The embodiments of the present application are further described below in conjunction with the accompanying drawings.

[0047] like Figure 1As shown, Figure 1 1 is a flow chart of a method for detecting internal defects of a municipal full water main provided by an embodiment of the present application. The method for detecting internal defects of a municipal full water main includes but is not limited to steps 110 to 160.

[0048] Step 110: Obtaining internal defect data of municipal full water mains;

[0049] Step 120: Characterize the internal defect data of the municipal full water main using a feature optimization strategy to obtain an internal defect feature map;

[0050] Step 130: constructing an initial full water main internal defect detection model;

[0051] Step 140: training the initial full water main internal defect detection model based on the internal defect feature map to obtain a trained full water main internal defect detection model;

[0052] Step 150: Obtain a characteristic diagram of internal defects to be detected;

[0053] Step 160: Input the internal defect feature map to be detected into the trained full water main internal defect detection model to perform defect detection and obtain the detection result.

[0054] Steps 110 to 160 are described in detail below.

[0055] In a feasible embodiment, in step 110, the internal defect data inside the full water main can be obtained by high-precision signal acquisition equipment (such as endoscopes, sonar detectors, etc.), and it can be understood that these devices can penetrate the water body, go deep into the pipe, and capture detailed information about the defects. The obtained raw data (internal defect data) usually exists in the form of a one-dimensional signal, which contains preliminary information about the defect.

[0056] In a feasible embodiment, in step 120, the original one-dimensional signal data obtained in step 110 is deeply characterized and converted through a feature optimization strategy, and can be converted into an intuitive and information-rich two-dimensional image (internal defect feature map). This process not only highlights the characteristics of the defects, but also greatly enhances the visibility and recognition of the defects, so that the subsequent detection model can more accurately capture and identify these defects, thereby significantly reducing the possible misjudgment and missed detection of the model during the detection process, and further improving the reliability and accuracy of the overall detection.

[0057] In a feasible embodiment, in step 130, an initial full water main internal defect detection model is designed and constructed based on the principle of deep learning. The model can be used to learn and identify defects inside the full water main. Effective model construction will improve the efficiency of subsequent training and the accuracy of detection results.

[0058] In a feasible embodiment, in step 140, the internal defect feature map obtained in step 120 is used as training data to train the initial detection model, and the model parameters are iteratively optimized so that the model can accurately identify the defects inside the full water main. It should be pointed out that the training process is the core link of improving the performance of the model, and sufficient training data and reasonable training strategies are required to ensure the generalization ability and accuracy of the model. Therefore, a municipal full water main internal defect database can be established through a signal acquisition device, and then the municipal full water main internal defect database can be characterized using a feature optimization strategy to obtain an internal defect feature library, and then the initial full water main internal defect detection model is trained based on the internal defect feature library to obtain a trained full water main internal defect detection model. It can be understood that the internal defect feature library has a large number of defect features, which can ensure that the model can accurately capture and learn a wider range of the most core defect information, and promote the comprehensive optimization of the detection model.

[0059] In a feasible embodiment, in the process of training the initial full water main internal defect detection model based on the internal defect feature map (internal defect feature library) to obtain the trained full water main internal defect detection model, the training set and the verification set can be first constructed according to the internal defect feature library; then the initial full water main internal defect detection model is trained and verified using the training set and the verification set, respectively, to obtain the full water main internal defect detection model. It should be pointed out that the ratio of the training set and the verification set can be divided into 70% and 30% of the total samples of the internal defect feature library, in order to ensure that the training set can fully cover and reflect the diversity of internal defects, while the verification set is used to objectively evaluate the generalization ability and detection accuracy of the model. It should be noted that the present application does not limit the ratio of the training set and the verification set, and the specific ratio can be set according to actual needs.

[0060] In a feasible embodiment, the full water main internal defect detection model is a trained machine learning model that can identify the internal defects of the full water main. For example, when it is necessary to identify the internal defects of the full water main, the corresponding defect detection results can be obtained by inputting the internal defect feature map of the full water main to be tested into the full water main internal defect detection model. Among them, the full water main internal defect detection model can be composed of a conventional deep neural network model or a convolutional neural network model, etc. For the relevant model structure and model principle of the deep neural network model or the convolutional neural network model, reference can be made to the introduction in the relevant technology, which will not be repeated here.

[0061] In a feasible embodiment, in step 160, internal defect data of the full water main to be tested can be obtained through a specific signal acquisition device (such as an endoscope, a sonar detector, etc.), and then these data are characterized through a feature optimization strategy to obtain a feature map of the internal defects to be detected. The trained full water main internal defect detection model is used to perform fast and accurate defect detection on the feature map of the internal defects to be detected, and the detection results are output.

[0062] In a feasible embodiment, in order to ensure the smooth progress of the detection process and optimize the detection effect, the input format of the internal defect feature map to be detected can be adjusted first. Specifically, the resolution of the image can be adjusted to 650 pixels x 650 pixels. This size selection is not fixed, but is flexibly adjusted according to actual detection requirements and model compatibility to achieve the best detection effect.

[0063] In a feasible embodiment, through the defect detection process from step 110 to step 160, the internal defect data of the municipal full water main is first obtained, and then the internal defect data of the municipal full water main is characterized by the feature optimization strategy to obtain the internal defect feature map. The internal defect data of the full water main is converted into the internal defect feature map by the feature optimization strategy. This process effectively extracts the key features of the defect, so that the subsequent detection model can more accurately identify and analyze the type of defect and other information, thereby improving the efficiency of data processing, and further enhancing the accuracy of the detection results. Further, an initial full water main internal defect detection model is constructed, and the initial full water main internal defect detection model is trained based on the internal defect feature map to obtain the full water main internal defect detection model. By inputting the acquired internal defect feature map to be detected into the trained full water main internal defect detection model for defect detection, the defect detection results can be obtained quickly and accurately, which greatly shortens the detection cycle, reduces the risk of human misjudgment, and provides strong technical support for the maintenance and management of municipal facilities. The embodiments of the present application solve the problems of low efficiency and poor accuracy faced by traditional detection methods in detecting internal defects of full water mains, thereby effectively improving the efficiency and accuracy of internal defect detection of full water mains.

[0064] See also Figure 2 In a feasible embodiment, the feature optimization strategy includes a time-frequency analysis algorithm, an adaptive window function and a search algorithm. In step 120, the feature optimization strategy is used to characterize the internal defect data of the municipal full water main pipe, and the processing flow of obtaining the internal defect feature map includes but is not limited to steps 210 to 230. :

[0065] Step 210: Perform time-frequency analysis and parameter extraction on the internal defect data using a time-frequency analysis algorithm to obtain a time-frequency aggregation degree;

[0066] Step 220: Obtain the standard deviation of the adaptive window function;

[0067] Step 230: Adaptively match the standard deviation of the adaptive window function according to the time-frequency concentration and the search algorithm until the standard deviation reaches a preset matching threshold, thereby obtaining an internal defect feature map.

[0068] It should be noted that the adaptive window function g(ut) and the time-frequency analysis algorithm G e The expressions of (t,ω), search algorithm ART(t,ω) and time-frequency aggregation degree CM are as follows:

[0069]

[0070]

[0071] Among them, x(u) is the collected signal, u is the time variable, j is the imaginary unit of the complex number, σ t is the standard deviation, ω is the frequency, t is the time, and D is the window length.

[0072] Steps 210 to 230 are described in detail below.

[0073] In a feasible embodiment, in step 210, the time-frequency analysis algorithm (such as short-time Fourier transform, wavelet transform, Hilbert-Huang transform, etc.) can provide information of the signal in the time and frequency domains at the same time, and is suitable for processing non-stationary signals or image data with time-varying characteristics. By performing time-frequency analysis on the internal defect data, important parameters reflecting the defect characteristics in the time-frequency distribution can be extracted, such as energy distribution, frequency change rate, etc. The time-frequency concentration can be used to measure the concentration or significance of the defect characteristics in the time-frequency domain, which helps to identify and strengthen these key features in subsequent steps.

[0074] In a feasible embodiment, in step 220, in image processing, the adaptive window function can be used to control the weight or shape of the data segment (or local area of ​​the image) to adapt to different analysis requirements. The adaptive window function can dynamically adjust its shape or size according to the characteristics of the data. The standard deviation can reflect the discreteness of the data. The standard deviation of the adaptive window function can be used to measure the variability or complexity of the data within the window function, thereby affecting the subsequent feature enhancement process.

[0075] In a feasible embodiment, in step 230, adaptive matching refers to iteratively adjusting the parameters (such as size, shape, position, etc.) of the adaptive window function using a search algorithm (such as gradient descent, genetic algorithm, particle swarm optimization, etc.), where the parameter is the standard deviation, so that the standard deviation and the time-frequency concentration are optimally matched. In general, the matching goal is to maximize the time-frequency concentration or to make the standard deviation meet a specific preset threshold, thereby strengthening the key defect features in the image.

[0076] In a feasible embodiment, the preset matching threshold can be a pre-set standard or limit for judging whether the adaptive matching process has achieved satisfactory results. When the standard deviation meets (such as being lower than or higher than) this threshold, it is considered that the internal defect feature has been sufficiently enhanced and the matching process ends.

[0077] In a feasible embodiment, step 120 (step 210 to step 230) effectively characterizes and enhances the internal defect data of the municipal full water main through time-frequency analysis, adaptive window function adjustment and optimization matching strategy, providing clearer and more accurate input data for subsequent defect detection.

[0078] In a feasible embodiment, the improved whale optimization algorithm (WOA) is integrated into the feature optimization strategy. By introducing the reverse learning strategy, the strategy combining Levy flight and Brownian motion, the global optimization capability of the WOA algorithm is not only enhanced, but also the feature optimization strategy can be more effectively focused on the extraction and enhancement of key defect features during implementation. This improvement not only accelerates the convergence speed of the optimization process, but also significantly improves the accuracy and reliability of the final internal defect feature map, laying a solid foundation for subsequent detection and identification tasks.

[0079] In a possible embodiment, the reverse learning strategy Levi Flying X i With Brownian motion X k The mathematical expression is:

[0080]

[0081] X i =X i-1 +α⊕levy(β),

[0082]

[0083] X k =X k-1 +rand(),

[0084] Among them, a represents the previous level of optimized parameters, b represents the next level of optimized parameters, and X irepresents the i-th individual in the population, α is the step factor, ⊕ represents the dot product, rand() represents the generation of a random number between 0 and 1, β=1.5, Γ is the standard Gamma function, and v obeys the standard normal distribution.

[0085] In a feasible embodiment, when using the time-frequency analysis algorithm to perform time-frequency analysis and parameter extraction on the internal defect data to obtain the time-frequency aggregation, the internal defect data can be first analyzed by the time-frequency analysis method to obtain the defect signal in the time-frequency domain; then the time-frequency aggregation is extracted from the defect signal in the time-frequency domain. Specifically, the time-frequency analysis algorithm is used to perform time-frequency analysis and parameter extraction on the internal defect data, and the defect features in the image can be converted into time-frequency domains in detail. This process not only reveals the complex distribution of the defect signal in the time and frequency dimensions, but also clearly outlines its unique form in the time-frequency space. Subsequently, the key parameter of time-frequency aggregation is accurately extracted from these defect signals in the time-frequency domain. It should be pointed out that the time-frequency aggregation can be used as an important indicator to measure the concentration and significance of defect features in the time-frequency domain, and its accurate extraction provides solid data support and decision-making basis for subsequent feature enhancement and identification work.

[0086] See also Figure 3 In a feasible embodiment, in step 160, the internal defect feature map to be detected is input into the full water main internal defect detection model for defect detection, and the processing flow of obtaining the detection results includes but is not limited to steps 310 to 330.

[0087] Step 310: performing feature processing on the internal defect feature map to be detected to obtain an enhanced feature map;

[0088] Step 320: reducing the dimension and transforming the enhanced feature map to obtain a reduced-dimensional feature map;

[0089] Step 330: Map and normalize the reduced-dimensional feature map to obtain a detection result.

[0090] Steps 310 to 330 are described in detail below.

[0091] In a feasible embodiment, step 310 is intended to pre-process the input internal defect feature map to be detected through a series of image processing or machine learning techniques (such as filtering, enhancement, edge detection, etc.) to highlight or enhance the key defect features in the image, while suppressing or removing unnecessary noise and interference. It can be understood that the enhanced feature map obtained after feature processing has more distinct defect features and is easier to be recognized and analyzed by the subsequent detection model.

[0092] In a feasible embodiment, in step 320, since the enhanced feature map may contain a large number of data dimensions, direct processing may lead to problems such as high computational complexity and long model training time. Therefore, by performing dimensionality reduction processing on it, the data dimension can be reduced, and the processing efficiency can be improved while retaining key information. In addition, the dimensionality reduction process is often accompanied by data conversion, that is, mapping the original high-dimensional data into a low-dimensional space, while trying to keep the intrinsic structure and key features of the data unchanged. The reduced dimensionality feature map obtained in this way is more compact and efficient, and is more suitable for subsequent detection model processing.

[0093] In a feasible embodiment, in step 330, mapping refers to mapping the reduced-dimensional feature map to the input space of the detection model (internal defect detection model of a full-water main), or a feature space conversion performed within the model to better match the detection logic and parameter settings of the model. Normalization is a data preprocessing technology that can adjust data of different dimensions or ranges to the same scale to avoid model training difficulties or performance degradation due to excessive data differences. During the detection process, normalizing the reduced-dimensional feature map can ensure the stability and consistency of the model input, thereby improving the accuracy and reliability of the detection results. The final output of the detection model may include information such as the location, type, and severity of the defect, providing an important reference for subsequent maintenance and repair work.

[0094] It can be understood that steps 310 to 330 constitute the core processing flow of inputting the internal defect feature image to be detected into the full water main internal defect detection model for defect detection. Through steps such as feature processing, dimensionality reduction transformation and normalization processing, the model can efficiently identify and analyze the defect features in the input image and output accurate detection results.

[0095] In a feasible embodiment, the internal defect detection model of the initial full water main can be an improved Densenet neural network model, including an improved DenseBlock module and a Transition module connected to the improved DenseBlock module. The improved DenseBlock module uses the Depthwise-Separable Convolution kernel to divide the original convolution work into two parts: Depthwise Convolution and Pointwise Convolution. First, Depthwise Convolution is used to use a convolution kernel for each channel of the input feature map, and then the outputs of all convolution kernels are spliced ​​to obtain its output, and then Pointwise Convolution is used to perform channel fusion on the feature map output by Depthwise Convolution.

[0096] In a practical implementation case, an optimized Densenet neural network model can be used as the core architecture for internal defect detection of an initially full water main. The model significantly improves the detection performance by introducing an improved DenseBlock module and a Transition module connected to the improved DenseBlock module. Specifically, these improved DenseBlock modules incorporate the Depthwise-Separable Convolution technology, creatively decomposing the traditional convolution operation into two efficient stages: First, the Depthwise Convolution stage, in which a convolution kernel is applied independently to each channel of the input feature map to achieve deep convolution processing of the feature map. This approach effectively reduces the amount of calculation while retaining the independent information between channels. Subsequently, all feature map outputs processed by the Depthwise Convolution are spliced ​​to form a richer feature set. Next, the Pointwise Convolution stage is followed, in which a 1x1 convolution kernel is used to fuse and transform the channel dimension of the feature map obtained by the previous step. This process not only realizes cross-channel information exchange, but also flexibly controls the dimension of the feature map by adjusting the number of output channels, which facilitates subsequent processing. Through the application of Depthwise-Separable Convolution, the improved DenseBlock module significantly reduces the computational complexity and number of parameters while maintaining powerful feature extraction capabilities, thereby improving the training efficiency and detection speed of the model. The Transition module, which is closely connected to these DenseBlock modules, is responsible for effective feature transfer and dimensionality adjustment between densely connected blocks, ensuring the smoothness and efficiency of the overall structure of the model.

[0097] See also Figure 4 , Figure 4It is the optimized Densenet network structure diagram provided by the embodiment of the present application. The Densenet neural network architecture starts with a 7x7 convolution kernel, which is responsible for the preliminary extraction of the basic features in the input image, followed by a Pooling layer for reducing the dimension of the feature map and enhancing the robustness of the feature. Subsequently, the network introduced an improved DenseBlock module, which is the core component of the entire architecture. Inside the improved DenseBlock, batch normalization (BN), ReLU activation function, 1x1 convolution kernel and 3x3 depth separable convolution (Depthwise-Separab le Conv) are integrated. Specifically, the feature is first normalized by the BN layer, and then the ReLU activation function is applied to introduce nonlinearity, and then the 1x1 convolution kernel is used to adjust the number of channels and feature fusion. Afterwards, the feature map enters the Depthwise-Separab le Conv stage, which first performs convolution operations on each channel independently through Depthwise Convo l ut ion, keeping the spatial dimension unchanged but increasing the depth of the feature map. Among them, each improved DenseBlock module is connected to a Transition module, which is responsible for reducing the size of the feature map and adjusting its number of channels to facilitate the processing of subsequent DenseBlock modules. This combination of DenseBlock and Transition modules is repeated three times in the network, forming a deep feature extraction network. It should be pointed out that after the last Transition module, a DenseBlock module is connected again to further integrate and strengthen the features extracted from all previous levels. Finally, the network outputs features through a SoftMax function. The role of the SoftMax layer is to convert the features extracted by the network into probability distribution, thereby realizing the classification and identification of internal defect types. This embodiment is based on the improved Densenet neural network structure by introducing advanced technologies such as deep separable convolution, batch normalization, ReLU activation, and a reasonable combination of DenseBlock and Transition modules to construct an efficient and powerful internal defect detection model for full water mains. Compared with classic classification network architectures, such as residual neural network (ResNet) and DenseNet, the internal defect detection model for full water mains proposed in this embodiment has achieved significant improvement in detection performance. It should be pointed out that while optimizing the network structure, the model greatly reduces the model complexity, effectively reducing the number of parameters from the original 6956931 parameters of DenseNet to 5123203, achieving a reduction of about 26% in the number of parameters, which is a significant reduction.In addition, the model significantly reduces the risk of missed detection and false detection, further enhancing its reliability and accuracy in practical applications.

[0098] In a feasible embodiment, the processing steps of performing feature processing on the feature map of the internal defect to be detected to obtain the enhanced feature map include: applying convolution kernels to each channel of the feature map of the internal defect to be detected, splicing the outputs of the convolution kernels to obtain the feature map; and performing channel fusion on the feature map to obtain the enhanced feature map. Specifically, firstly, the convolution kernel is applied to each channel of the feature map of the internal defect to be detected, so that the key defect features in each channel can be captured independently while maintaining the independence of the information between the channels. Subsequently, the output results of all convolution kernels are spliced ​​to form a preliminary feature map containing rich multi-channel feature information. The preliminary feature map not only retains the spatial structure of the original image (the feature map of the internal defect to be detected), but also significantly enhances the defect details in the feature map. In order to further enhance the representation ability of the feature map, channel fusion processing can also be performed on it. It should be noted that channel fusion is an efficient information integration technology that allows the model to learn and utilize the relationship between features across channels. Through this step, the useful information scattered in each channel in the preliminary feature map can be effectively integrated to generate an enhanced feature map. This enhanced feature map not only makes the defect features more distinct, but also makes the overall information more compact and efficient, providing better quality input data for subsequent detection models.

[0099] In a feasible embodiment, after obtaining the test results, the method also includes: when the test results meet the preset results, using the test results as output results; when the test results are judged to fail to meet the preset results, correcting the weight of the internal defect detection model of the full water main.

[0100] In a practical implementation case, after completing the internal defect detection and obtaining preliminary results, a subsequent processing strategy can also be adopted for the detection results. The processing strategy includes: when the detection result meets the preset result, the detection result is used as the output result; when the detection result is judged to fail to meet the preset result, the weight of the internal defect detection model of the full water main is corrected. Specifically, the detection result is first evaluated to determine whether it meets the preset standard or threshold (i.e., the preset result). If the detection result meets these requirements, it proves that the current detection model performs well and can accurately identify the defects inside the full water main. This detection result can be directly used as the final output result for subsequent analysis, decision-making or maintenance processes. However, if the detection result fails to meet the preset standard, it means that the current detection model may be insufficient or biased in identifying specific defects. In order to improve the accuracy and robustness of the model, an active feedback adjustment mechanism can be adopted. Specifically, the weight of the internal defect detection model of the full water main will be corrected according to the wrong detection result. This process may involve fine-tuning the model parameters, adjusting the learning rate, or introducing more training samples to cover the defect types that have not been fully learned. Through continuous iteration and optimization, the aim is to make the detection model better adapt to the needs of actual detection scenarios and improve the accuracy and reliability of detection. It should be noted that after the accuracy of the full water main internal defect detection model has reached 95% through the verification set, the detection results obtained by using the full water main internal defect detection model to perform defect detection on the defect image to be tested can be regarded as reaching the preset results.

[0101] Reference Figure 5 The embodiment of the present application further discloses an electronic device, the electronic device 500 comprising:

[0102] at least one processor 501;

[0103] At least one memory 502, used to store at least one program;

[0104] When at least one program is executed by at least one processor 501, the above-mentioned method for detecting internal defects of a municipal full water main is implemented.

[0105] An embodiment of the present application also discloses a computer-readable storage medium, which stores a computer program executable by a processor. When the computer program executable by the processor is executed by the processor, it is used to implement the internal defect detection method of a municipal full water main as described above.

[0106] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting internal defects of a municipal full water main, characterized in that: include: Obtain internal defect data of municipal full water mains; Characterizing the internal defect data of the municipal full water main using a feature optimization strategy to obtain an internal defect feature map; Construct an internal defect detection model for an initially full water main; Training the initial full water main internal defect detection model based on the internal defect feature map to obtain a trained full water main internal defect detection model; Obtaining a characteristic diagram of internal defects to be detected; The internal defect feature map to be detected is input into the trained full water main internal defect detection model to perform defect detection and obtain a detection result.

2. The method for detecting internal defects of a municipal full water main according to claim 1, characterized in that: The feature optimization strategy includes a time-frequency analysis algorithm, an adaptive window function and a search algorithm. The feature optimization strategy is used to characterize the internal defect data of the municipal full water main to obtain an internal defect feature map, including: Performing time-frequency analysis and parameter extraction on the internal defect data using a time-frequency analysis algorithm to obtain a time-frequency aggregation degree; Obtaining a standard deviation of the adaptive window function; Adaptively matching the standard deviation of the adaptive window function according to the time-frequency concentration and the search algorithm until the standard deviation reaches a preset matching threshold, thereby obtaining an internal defect feature map; Among them, the adaptive window function g(ut), the time-frequency analysis algorithm G e (t,ω), the search algorithm ART(t,ω) and the expression of time-frequency aggregation are: Among them, x(u) is the collected signal, u is the time variable, j is the imaginary unit of the complex number, σ t is the standard deviation, ω is the frequency, t is the time, and D is the window length.

3. The method for detecting internal defects of a municipal full water main according to claim 2, characterized in that: The method of using a time-frequency analysis algorithm to perform time-frequency analysis and parameter extraction on the internal defect data to obtain a time-frequency aggregation degree includes: using a time-frequency analysis method to analyze the internal defect data to obtain a defect signal in the time-frequency domain; and extracting the time-frequency aggregation degree from the defect signal in the time-frequency domain.

4. The method for detecting internal defects of a municipal full water main according to claim 1, characterized in that: The method of inputting the internal defect feature map to be detected into the trained full water main internal defect detection model to perform defect detection and obtain a detection result includes: performing feature processing on the internal defect feature map to be detected to obtain an enhanced feature map; performing dimensionality reduction and transformation on the enhanced feature map to obtain a reduced dimensionality feature map; and performing mapping and normalization processing on the reduced dimensionality feature map to obtain a detection result.

5. According to the municipal full water main internal defect detection method of claim 4, the feature processing of the internal defect feature map to be detected to obtain an enhanced feature map comprises: Using convolution kernels on each channel of the internal defect feature map to be detected, and splicing the outputs of the convolution kernels to obtain a feature map; Channel fusion is performed on the feature map to obtain an enhanced feature map.

6. The method for detecting internal defects of a municipal full water main according to claim 1, further comprising: When the detection result meets the preset result, the detection result is used as the output result; When the detection result fails to meet the preset result, the weight of the internal defect detection model of the full water main is corrected.

7. According to the municipal full water main internal defect detection method of claim 2, the feature optimization strategy includes a reverse learning strategy Levi Flying X i With Brownian motion X k , reverse learning strategy Levi Flying X i With Brownian motion X k The mathematical expression is: XkXk-1+rand0 in, a represents the previous level of optimized parameters, b represents the next level of optimized parameters, and X i represents the i-th individual in the population, α is the step size factor, represents the dot product, rand() represents the generation of a random number between 0 and 1, β=1.5, Γ is the standard Gamma function, and v follows a standard normal distribution.

8. The method for detecting internal defects of a municipal full water main according to claim 1, characterized in that: The method of training the initial full water main internal defect detection model based on the internal defect feature map to obtain a trained full water main internal defect detection model includes: constructing a training set and a validation set according to the internal defect feature map; and using the training set and the validation set to respectively train and validate the initial full water main internal defect detection model to obtain the full water main internal defect detection model.

9. An electronic device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When at least one of the programs is executed by at least one of the processors, the method for detecting internal defects of a municipal full water main as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program executable by a processor is stored therein, and when the computer program executable by the processor is executed by the processor, it is used to implement the method for detecting internal defects of a municipal full water main as described in any one of claims 1 to 8.

Citation Information

Patent Citations

  • Pipe leak detection method and device

    CN109931506A

  • Drainage pipeline working condition detection and identification method

    CN112198232A

  • Municipal drainage pipeline defect detection system and method thereof

    CN116721051A

  • Multi-domain feature extraction and fusion identification method for water supply pipeline leakage

    CN116753471A

  • Pipeline detection method and system, terminal equipment and storage medium

    CN117173130A