Urban drainage pipe network functional defect automatic identification and maintenance grading method and system

Abstract

The invention relates to the technical field of image processing, and discloses an urban drainage pipe network functional defect automatic identification and maintenance grading method and system. The method comprises the following steps: acquiring a drainage pipe network video through pipeline detection equipment, and performing coding extraction according to six-dimensional information such as city and the like to obtain a defect picture; carrying out decoding and feature extraction on the picture to obtain a feature matrix; defect type data is formed through key defect screening, feature refinement and parameter quantification; performing grade evaluation and classification to obtain an evaluation result; and finally, matching and sequencing the maintenance schemes to generate a maintenance decision scheme. According to the invention, automatic identification and grading of defects are realized, and a scientific maintenance decision scheme is provided, so that the efficiency and accuracy of pipe network maintenance work are improved.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a method and system for automatically identifying and grading the maintenance of functional defects in urban drainage networks. Background Art

[0002] At present, as an important part of urban infrastructure, the operation status of urban drainage pipe networks is directly related to urban flood control and drainage and the quality of life of residents. With the acceleration of urbanization, the scale of drainage pipe networks continues to expand, and the problem of pipe network defects has become increasingly prominent. The existing drainage pipe network detection mainly relies on CCTV detection equipment for video acquisition, and then manually watching the video to identify the type and level of defects. At the same time, the industry has developed a variety of pipeline detection equipment and image processing technologies, such as sonar detection and laser scanning, which can obtain information on the internal status of the pipeline from different angles. In terms of defect assessment, various regions have established corresponding pipe network assessment standards and maintenance specifications, which provide a basic basis for pipe network maintenance work.

[0003] However, the existing technology has the following problems: First, the manual defect identification method is inefficient and easily affected by subjective factors, resulting in inaccurate and inconsistent evaluation results; second, the data formats collected by various types of detection equipment are not unified, and there is a lack of standardized data management mechanisms, which makes it difficult to effectively use the data; third, the defect assessment process often only considers a single feature, ignoring the multi-dimensional feature information of the defect, affecting the comprehensiveness of the evaluation results; fourth, maintenance decisions mainly rely on experience and judgment, lack of a scientific priority sorting mechanism, and it is difficult to achieve the optimal allocation of maintenance resources. Summary of the invention

[0004] The present application provides a method and system for automatically identifying and grading functional defects in an urban drainage network, which is used to realize automatic identification and grading of defects and provide a scientific maintenance decision-making plan to improve the efficiency and accuracy of network maintenance work.

[0005] In the first aspect, the present application provides a method for automatically identifying and grading the functional defects of an urban drainage network, the method comprising: obtaining a drainage network detection video through a pipeline detection device, encoding and extracting the detection video according to six-dimensional information to obtain network defect picture data, wherein the six-dimensional information includes city, region, year, defect name, grade and serial number; performing image decoding and feature extraction on the network defect picture data to obtain a pipeline defect feature matrix; performing key defect screening according to the pipeline defect feature matrix to obtain a key defect picture set. , including: setting a discrimination threshold after normalizing the values ​​in the pipeline defect feature matrix, obtaining a defect focus area by extracting super-threshold feature points and performing spatial clustering analysis; calculating a hazard score according to the area and depth of the defect focus area, screening out defects with high hazard levels and extracting corresponding original images to obtain a key defect image set; performing feature refinement and parameter quantification processing on the key defect image set to obtain defect type data; performing grade evaluation and classification processing on the defect type data to obtain a defect grade evaluation result; performing maintenance plan matching and priority sorting processing according to the defect grade evaluation result to obtain a drainage network maintenance decision plan.

[0006] In a first implementation of the first aspect, a drainage network inspection video is obtained through a pipeline inspection device, and the inspection video is encoded and extracted according to six-dimensional information to obtain network defect image data, wherein the six-dimensional information includes city, region, year, defect name, level and serial number, including: video acquisition of the drainage network through a pipeline CCTV inspection device, and frame extraction of the drainage network inspection video according to a preset sampling frequency; denoising and illumination compensation preprocessing of the frame image sequence of the inspection video to remove image noise and uneven illumination effects; extracting key frame images from the preprocessed frame image sequence, and establishing a spatiotemporal index for the key frame images; encoding and identifying the key frame images according to the city, region, year, defect name, level and serial number in the six-dimensional information; associating and matching the encoded identifier with the corresponding key frame image to generate image data with a unique identifier; batch sorting and format conversion of the image data with a unique identifier to form the network defect image data.

[0007] In a second implementation of the first aspect, the image decoding and feature extraction processing of the pipeline network defect picture data to obtain a pipeline defect feature matrix includes: performing pixel point distribution statistical analysis on the pipeline network defect picture data to obtain image grayscale distribution data; performing threshold segmentation processing on the image grayscale distribution data to obtain a binary image of the defect area; performing edge detection and contour extraction on the binary image of the defect area to obtain a defect boundary point set; performing geometric feature calculation on the defect boundary point set to extract defect area, perimeter, and roundness measurement values; performing texture feature analysis on the defect area to extract energy, contrast, and correlation parameters of the grayscale co-occurrence matrix; and performing data fusion on the geometric features and texture features to obtain the pipeline defect feature matrix.

[0008] In a third implementation of the first aspect, the key defect screening process is performed according to the pipeline defect feature matrix to obtain a key defect picture set, including: normalizing the characteristic values ​​in the pipeline defect feature matrix to obtain standardized characteristic data; setting a defect discrimination threshold for the standardized characteristic data, and extracting characteristic points that exceed the discrimination threshold; performing spatial clustering analysis on the characteristic points that exceed the discrimination threshold to obtain a defect focus area; calculating the area and depth of the defect focus area to obtain defect degree data; calculating a hazard score according to the defect degree data to screen out high hazard degree defect marks; extracting and sorting the original pictures corresponding to the high hazard degree defect marks to obtain the key defect picture set.

[0009] In a fourth embodiment of the first aspect, the key defect picture set is subjected to feature refinement and parameter quantification processing to obtain defect type data, including: performing pixel gradient calculation on the key defect picture set to obtain defect edge contour lines; performing shape feature analysis on the defect edge contour lines to extract defect morphological parameters; performing numerical segmented interval statistics on the defect morphological parameters to obtain defect type feature vectors; associating and matching the defect type feature vectors with defect location information to form spatial distribution data; performing defect degree score calculation on the spatial distribution data to generate a defect severity index; combining the defect severity index with the corresponding morphological features to obtain the defect type data.

[0010] In a fifth implementation of the first aspect, the defect type data is graded and classified to obtain a defect grade assessment result, including: characteristically grouping the defect type data according to defect morphology to obtain a defect category statistical table; performing weight assignment calculation on the data in the defect category statistical table to obtain a defect impact factor; correlating the defect impact factor with pipeline operation status data to obtain a defect hazard index; dividing the defect hazard index into graded intervals to generate a defect grade determination standard; comparing and verifying the defect grade determination standard with historical maintenance records to form a defect grade mapping table; and performing data integration and grade matching on the defect grade mapping table to obtain the defect grade assessment result.

[0011] In a sixth implementation of the first aspect, the maintenance scheme matching and priority sorting processing are performed according to the defect level assessment results to obtain a drainage network maintenance decision plan, including: comparing the defect level assessment results with the maintenance technology scheme library to form a maintenance measure correspondence table; cost accounting for the maintenance schemes in the maintenance measure correspondence table to obtain a maintenance resource demand list; priority calculation of the maintenance resource demand list according to the severity of the defects to generate a maintenance schedule; construction condition constraint analysis of the maintenance schedule to form a feasibility assessment report; matching and analyzing the feasibility assessment report with existing maintenance resources to obtain a maintenance scheduling plan; and comprehensively optimizing and integrating the maintenance scheduling plan to obtain the drainage network maintenance decision plan.

[0012] In a second aspect, the present application provides an automatic identification and maintenance grading system for functional defects of an urban drainage network, the automatic identification and maintenance grading system for functional defects of an urban drainage network comprising: The acquisition module is used to obtain the drainage network inspection video through the pipeline inspection equipment, encode and extract the inspection video according to the six-dimensional information, and obtain the pipeline network defect image data, wherein the six-dimensional information includes city, region, year, defect name, level and serial number; An extraction module is used to perform image decoding and feature extraction processing on the pipeline network defect image data to obtain a pipeline defect feature matrix; A screening module is used to screen key defects according to the pipeline defect feature matrix to obtain a key defect picture set, including: setting a discrimination threshold after normalizing the values ​​in the pipeline defect feature matrix, extracting super-threshold feature points and performing spatial clustering analysis to obtain a defect focus area; calculating a hazard score according to the area and depth of the defect focus area, screening out defects with high hazard levels and extracting corresponding original pictures to obtain a key defect picture set; A quantification module is used to perform feature refinement and parameter quantification processing on the key defect image set to obtain defect type data; A classification module, used to perform grade assessment and classification processing on the defect type data to obtain a defect grade assessment result; The sorting module is used to match and prioritize maintenance plans according to the defect level assessment results to obtain a drainage network maintenance decision plan.

[0013] In the technical solution provided by the present application, the drainage network inspection video is obtained through the pipeline inspection equipment, and the encoding and extraction processing is performed based on the six-dimensional information, a standardized defect data collection and management mechanism is established, and the standardized storage and efficient retrieval of the inspection data are realized, laying a data foundation for the subsequent intelligent analysis; the feature matrix of the pipeline defects is obtained by image decoding and feature extraction processing of the pipeline network defect picture data, so that the defect feature extraction process is more objective and comprehensive, and the subjectivity and inconsistency of manual interpretation are avoided; the pipeline defect feature matrix is ​​screened for key defects, which improves the recognition and attention of important defects and effectively prevents important defects from being omitted or ignored; the key defect picture set is subjected to feature refinement and parameter quantification processing, which realizes the accurate quantification of defect features and provides reliable data support for defect assessment; the defect type data is graded and classified, and a scientific defect grading system is established, making the defect grade determination more accurate and objective; the maintenance plan is matched and prioritized according to the defect grade assessment results, which realizes the scientific and automated maintenance decision-making, improves the rationality and implementation effect of the maintenance plan, and greatly improves the work efficiency and accuracy of drainage network defect identification and maintenance management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0015] Figure 1 A schematic diagram of an embodiment of a method for automatically identifying and grading functional defects of an urban drainage network in an embodiment of the present application; Figure 2 This is a schematic diagram of six-dimensional information in an embodiment of the present application; Figure 3 This is a schematic diagram of an embodiment of the automatic identification and maintenance classification system of functional defects in urban drainage pipe networks in the embodiments of the present application. DETAILED DESCRIPTION

[0016] The embodiment of the present application provides a method and system for automatically identifying and grading the functional defects of an urban drainage network. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application 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 understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks includes: Step S101: Obtain a drainage network inspection video through a pipeline inspection device, encode and extract the inspection video according to six-dimensional information, and obtain network defect image data, wherein the six-dimensional information includes city, region, year, defect name, level, and serial number; Step S102, performing image decoding and feature extraction processing on the pipeline network defect image data to obtain a pipeline defect feature matrix; Step S103, performing key defect screening processing according to the pipeline defect feature matrix to obtain a key defect picture set, including: normalizing the values ​​in the pipeline defect feature matrix and setting a discrimination threshold, extracting super-threshold feature points and performing spatial clustering analysis to obtain a defect focus area; calculating a hazard score according to the area and depth of the defect focus area, screening out defects with high hazard levels and extracting corresponding original pictures to obtain a key defect picture set; Step S104, performing feature refinement and parameter quantification processing on the key defect picture set to obtain defect type data; Step S105, performing grade assessment and classification processing on the defect type data to obtain a defect grade assessment result; Step S106: Match and prioritize maintenance plans based on defect level assessment results to obtain a drainage network maintenance decision plan.

[0018] It is understandable that the execution subject of the present application may be the automatic identification and maintenance classification system for functional defects of urban drainage pipe networks, or may be a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking the server as the execution subject as an example.

[0019] Specifically, the video data inside the pipeline is obtained through CCTV pipeline detection equipment. The pipeline detection equipment is equipped with a high-definition camera to perform a 360-degree scan of the inner wall of the pipeline without blind spots, and collect video data with a resolution of 1920×1080 pixels. In the process of video data collection, six-dimensional information is used for encoding and identification, such as Figure 2 , which is a schematic diagram of six-dimensional information in an embodiment of the present application, Figure 2 The structural diagram of the six-dimensional information coding system of drainage network defects is shown. The coding system consists of six dimensions: city, region, year, defect name, grade and serial number. The city dimension is represented by capital letter abbreviations (such as "SH"), the region dimension is also represented by capital letter abbreviations (such as "PD"), the year dimension uses four digits to represent the inspection year (such as "2017"), the defect name dimension uses English letter combinations to represent different types of defects (such as "CJ"), the grade dimension uses two-digit codes to represent the severity of the defect (such as "03"), and the serial number dimension uses six digits for sequential numbering (such as "000440"). The figure shows the relationship between these six dimensions and the core concepts in a radial layout, and gives a complete coding example "SH PD 2017 CJ 03 000440" at the bottom, which intuitively shows the combination of information in each dimension, and provides a systematic coding specification for the standardized management and efficient retrieval of drainage network defect data. The city dimension in the six-dimensional information is represented by capital letter abbreviations, such as "SH" for a city; "PD" for a region; the year is represented by four digits; the defect name uses a combination of letters such as "CJ" to represent different types of defects; the grade is coded with two digits; the serial number is numbered in sequence with six digits. Based on this coding rule, a complete coding example of a pipeline inspection video is "SH PD 2017 CJ 03000440". The collected video data is frame extracted, and the 30fps original video is converted into a picture sequence at a frequency of extracting 1 frame per second. Then, through image noise reduction and illumination compensation, irregular light spots and shadow interference are eliminated, and finally standardized pipeline network defect picture data is generated. When decoding and extracting features from the acquired pipeline network defect picture data, the pixel grayscale distribution of the image is counted, and the grayscale histogram data of the image is calculated. The image is converted into a binary image by setting a grayscale threshold, and the regional contour of the pipeline defect is extracted. The defect area is edge detected, the boundary point set sequence is extracted, and the geometric feature parameters of the defect are calculated, including the defect area, perimeter, and roundness measurement value. At the same time, gray-level co-occurrence matrix analysis is performed on the defect area to extract texture feature parameters such as energy, contrast, and correlation. All geometric feature parameters and texture feature parameters are organized into feature vectors to construct a pipeline defect feature matrix. Each row in the feature matrix corresponds to a defect sample, and each column corresponds to a feature dimension.

[0020] The data in the pipeline defect feature matrix are normalized, and the feature data of different dimensions are uniformly mapped to the interval [0, 1]. The feature discrimination threshold is set according to the prior knowledge of the defect type, the feature data is screened, and the abnormal feature points exceeding the threshold are extracted. The screened feature points are subjected to spatial clustering analysis, and adjacent feature points are classified as the same defect focus area. The area and depth index of each defect focus area are calculated, and the hazard degree of the defect is evaluated based on the product of the area and depth. High-risk defects are screened according to the hazard degree score results, and the corresponding original defect images are extracted and sorted to form a key defect image set. The pixel gradient is calculated for each image in the key defect image set to extract the edge contour line of the defect. The shape characteristics of the contour line are analyzed, including the morphological parameters such as the perimeter, area, roundness, and eccentricity of the contour. The morphological parameters are subjected to numerical interval statistics to construct the feature vector of the defect type. The feature vector is associated with the spatial location information of the defect in the pipeline to form the spatial distribution data of the defect. The defect severity index is calculated according to the significance of the morphological characteristics, and the severity index is combined with the morphological feature data to generate standardized defect type data.

[0021] The defect type data is grouped and counted according to the morphological characteristics to establish a defect category frequency table. Weight coefficients are assigned to different types of defects, and the influencing factors of various defects are calculated. The influencing factors are correlated with the actual operation status data of the pipeline to evaluate the degree of harm of the defects to the pipeline operation. The grading interval is set according to the harm index, and the defects are divided into different levels. The rationality of the grading results is verified by comparing the historical maintenance records, and a mapping relationship table of defect levels is established, and the defect level assessment results are finally output. According to the defect level assessment results, the maintenance technology solution library is queried to match the corresponding maintenance measures. Cost accounting is performed on the maintenance plan, including labor costs, equipment costs, material costs, etc., to generate a detailed resource demand list. Prioritize the maintenance tasks according to the severity of the defects and formulate a preliminary maintenance schedule. Analyze the constraints such as construction site conditions, weather factors, and traffic control to evaluate the feasibility of the maintenance plan. Match the feasible maintenance plan with the existing human and material resources, reasonably allocate maintenance resources, and formulate a maintenance scheduling plan. Comprehensively consider multiple factors to optimize and adjust the maintenance plan to form a drainage network maintenance decision plan.

[0022] In the embodiment of the present application, the drainage network detection video is obtained by the pipeline detection equipment, and the encoding and extraction processing is performed according to the six-dimensional information, a standardized defect data collection and management mechanism is established, and the standardized storage and efficient retrieval of the detection data are realized, which lays a data foundation for the subsequent intelligent analysis; the feature matrix of the pipeline defect is obtained by image decoding and feature extraction processing of the pipeline network defect picture data, so that the defect feature extraction process is more objective and comprehensive, and the subjectivity and inconsistency of manual interpretation are avoided; the pipeline defect feature matrix is ​​screened for key defects, which improves the recognition and attention of important defects and effectively prevents important defects from being omitted or ignored; the key defect picture set is subjected to feature refinement and parameter quantification processing, which realizes the accurate quantification of defect features and provides reliable data support for defect assessment; the defect type data is graded and classified, and a scientific defect grading system is established, so that the defect grade determination is more accurate and objective; the maintenance plan matching and priority sorting processing are performed according to the defect grade assessment results, which realizes the scientific and automated maintenance decision-making, improves the rationality and implementation effect of the maintenance plan, and greatly improves the work efficiency and accuracy of drainage network defect identification and maintenance management.

[0023] In a specific embodiment, the process of executing step S101 may specifically include the following steps: (1) Capture video of the drainage network through pipeline CCTV detection equipment, and extract frames from the drainage network detection video according to a preset sampling frequency; (2) Perform noise reduction and illumination compensation preprocessing on the frame image sequence of the detection video to remove image noise and the influence of uneven illumination; (3) Extract key frame images from the preprocessed frame image sequence and establish spatiotemporal indexes for the key frame images; (4) Encode and identify the key frame images according to the city, region, year, defect name, grade, and serial number in the six-dimensional information; (5) Associating and matching the coding identifier with the corresponding key frame image to generate image data with a unique identifier; (6) Batch organize and convert the image data with unique identifiers to form pipeline network defect image data.

[0024] Specifically, when the drainage network is video-collected by the pipeline CCTV detection equipment, the CCTV detection equipment is a professional pipeline endoscopic detection equipment, which is mainly composed of a camera, a lighting system, a walking mechanism and a data transmission unit. When the drainage network is inspected, the CCTV detection equipment moves at a uniform speed along the pipeline axis, and the camera collects image information of the inner wall of the pipeline in real time. Based on the video frame rate of 30fps (30 frames per second) collected by the equipment, considering the spatiotemporal continuity characteristics of pipeline defects, the preset sampling frequency is set to 5fps, that is, 5 frames of images are extracted from the original video per second. The selection of sampling frequency requires a trade-off between data volume and information integrity. Too high sampling frequency will lead to data redundancy, and too low sampling frequency may lose key information. After frame extraction processing, the original video is converted into a discrete image sequence. For the extracted frame image sequence, noise reduction and illumination compensation preprocessing are performed. The noise reduction process uses a median filtering algorithm. By sliding a fixed-size window (such as 3×3 or 5×5) on the image, the central pixel value is replaced by the median of the pixel value in the window, effectively removing salt and pepper noise. The illumination compensation adopts the histogram equalization method to improve the image contrast by redistributing the gray value distribution of the image. For the common local over-darkness or over-brightness phenomenon in pipeline images, adaptive histogram equalization is used to process the image in blocks, and each block is histogram equalized separately, and then the boundaries between blocks are smoothed using the bilinear interpolation method to achieve a more uniform illumination effect.

[0025] When extracting key frames from the preprocessed frame image sequence, a content-based key frame extraction algorithm is used. By calculating the difference between adjacent frames, including color histogram difference, texture feature difference and edge feature difference, when the difference exceeds the preset threshold, the frame is marked as a key frame. A spatiotemporal index is established for the extracted key frames, and the index structure contains timestamps, spatial location information (including pipe section number, mileage pile number) and image feature descriptors. The spatiotemporal index is stored in a B+ tree structure to support efficient time and space range queries. When encoding and identifying key frame images with six-dimensional information, the encoding rules strictly follow the format of "city-region-year-defect name-level-serial number". The city is abbreviated with two capital letters, the region is abbreviated with two capital letters, the year is a four-digit number, the defect name is abbreviated with two capital letters (such as CJ for deposition and FS for corrosion), the level is a two-digit number (01-04 for level one to four), and the serial number is a six-digit number. During the encoding process, each dimensional information is normalized to ensure the uniqueness and readability of the encoding. Each key frame image will be assigned a unique encoding identifier.

[0026] When associating and matching the coded identifier with the key frame image, a key-value pair storage structure is used, with the coded identifier as the key and the image file path and metadata as the value. During the association matching process, a metadata file is generated for each key frame image, which contains information such as image acquisition time, spatial location, and equipment parameters. The metadata is stored in JSON format to facilitate data serialization and deserialization operations. After the matching is completed, each image data has a unique identifier to facilitate subsequent retrieval and processing. When batch sorting and format conversion of image data with unique identifiers, the images are uniformly converted to a specified format (such as PNG format) and appropriately compressed. Image compression uses a lossless compression algorithm to ensure that image detail information is not lost. During the data sorting process, the images are sorted according to time series and spatial location, and a hierarchical directory structure is established for storage. The format conversion and file organization work are automated through batch scripts to generate a standardized pipeline network defect image dataset.

[0027] For example, when CCTV inspection of drainage pipe network in a certain area of ​​a city is carried out, the original video acquisition time is 120 minutes, the video frame rate is 30fps, and the resolution is 1920×1080. The frame image is extracted at a sampling frequency of 5fps to obtain a sequence of 36,000 frames. In the preprocessing stage, a 5×5 median filter window is used for noise reduction, and the image is divided into 4×4 blocks for processing during adaptive histogram equalization. When extracting key frames, the color histogram difference threshold is set to 0.3, the texture feature difference threshold is set to 0.25, and the edge feature difference threshold is set to 0.2, and finally 3,600 key frames are extracted. Each key frame is given a unique coding identifier, such as "AB CD 2023 CJ 03 000001" represents the No. 1 third-level sediment defect detected in the area in 2023. In the association matching stage, a JSON format metadata file containing information such as image path, acquisition time, and pipe section number is generated. Finally, all images are converted into PNG format and organized and stored according to the directory structure of "city / region / year / defect type" to form a defect image dataset.

[0028] In a specific embodiment, the process of executing step S102 may specifically include the following steps: (1) Perform pixel distribution statistical analysis on pipeline network defect image data to obtain image grayscale distribution data; (2) The image grayscale distribution data is processed by threshold segmentation to obtain a binary image of the defect area; (3) Perform edge detection and contour extraction on the binary image of the defect area to obtain a set of defect boundary points; (4) Calculate the geometric features of the defect boundary point set and extract the defect area, circumference, and roundness measurement values; (5) Analyze the texture features of the defect area and extract the energy, contrast, and correlation parameters of the gray-level co-occurrence matrix; (6) The geometric features and texture features are fused to obtain the pipeline defect feature matrix.

[0029] Specifically, when performing pixel distribution statistical analysis on the pipeline network defect image data, the grayscale value of each pixel is counted, and the grayscale value range is 0-255. The grayscale histogram is obtained by counting the frequency of occurrence of each grayscale value. The grayscale histogram reflects the distribution of different grayscale levels in the image. By normalizing the histogram, the frequency is converted into a probability distribution. For each grayscale level, its cumulative distribution function is calculated. When the grayscale distribution data is processed by threshold segmentation, the Otsu algorithm is used to adaptively determine the optimal threshold. The Otsu algorithm divides the grayscale of the image into two categories by the maximum inter-class variance criterion, so that the inter-class variance is maximized. In the specific calculation, all grayscale levels are traversed as candidate thresholds, the inter-class variance is calculated, and the grayscale value that maximizes the inter-class variance is selected as the segmentation threshold. According to the obtained threshold, the pixels in the image are divided into foreground (defective area) and background, and a binary image is generated, in which the pixel value of the defective area is 1 and the pixel value of the background area is 0.

[0030] When performing edge detection on the binary image of the defect area, the Canny operator is used for processing. The image is subjected to Gaussian smoothing filtering to remove the influence of noise. Then the gradient amplitude and direction of the image are calculated, and the local maximum gradient value point is retained through non-maximum suppression. Finally, the edge is obtained through double threshold detection and connection analysis. After obtaining the edge, the boundary point set of the defect area is extracted using the contour tracking algorithm. The boundary points are arranged in a clockwise direction, and each point contains its coordinate information.

[0031] For the calculation of the geometric features of the defect boundary point set, the following formula is used: Defect area calculation formula: ; in: is the defect area; n is the number of boundary points; ( , ) is the coordinate of the i-th boundary point; ( , ) are the coordinates of the i+1th boundary point.

[0032] Defect perimeter calculation formula: ; in: is the defect perimeter.

[0033] The roundness measurement value calculation formula is: ; in: is the roundness measurement value; is the distance from the i-th boundary point to the defect center; is the maximum distance from all boundary points to the center.

[0034] When analyzing the texture features of defective areas, the texture features are characterized by calculating the grayscale co-occurrence matrix. The grayscale co-occurrence matrix describes the spatial correlation of grayscale values ​​in an image, usually considering four directions: 0°, 45°, 90°, and 135°. For each direction, the frequency of occurrence of grayscale values ​​of pixel pairs with a specific spatial relationship is counted to form a co-occurrence matrix. Three characteristic parameters, energy, contrast, and correlation, are extracted from the co-occurrence matrix. Energy reflects the uniformity of the grayscale distribution of the image, contrast represents the clarity of the image, and correlation describes the degree of correlation of local areas of the image.

[0035] When fusing geometric features and texture features, feature vector concatenation is used. Each feature is normalized to unify the numerical range to the [0,1] interval. Then the normalized geometric features (area, perimeter, roundness) and texture features (energy, contrast, correlation) are combined into a feature vector in a fixed order. For each defect image, a 6-dimensional feature vector is generated. Multiple feature vectors form a feature matrix. Each row of the matrix corresponds to a defect sample, and each column corresponds to a feature dimension.

[0036] For example: for a pipeline defect image, its grayscale distribution is counted, and the optimal threshold of 147 is calculated by the Otsu algorithm to divide the image into defect area and background. The Canny operator is used for edge detection to obtain the defect boundary and extract 384 boundary points. The geometric features of the defect are calculated based on the boundary points: the area is 12,800 square pixels, the perimeter is 640 pixels, and the roundness measurement value is 0.78. The grayscale co-occurrence matrix is ​​calculated for the defect area, and the energy value in the 0° direction is 0.65, the contrast is 0.42, and the correlation is 0.83. After normalizing all features, the feature vector [0.75, 0.82, 0.78, 0.65, 0.42, 0.83] is obtained. This completes the conversion process from the original image to the feature matrix.

[0037] In a specific embodiment, the process of executing step S103 may specifically include the following steps: (1) Normalize the characteristic values ​​in the pipeline defect characteristic matrix to obtain standardized characteristic data; (2) Setting a defect discrimination threshold for the standardized feature data and extracting feature points that exceed the discrimination threshold; (3) Perform spatial cluster analysis on the feature points that exceed the discrimination threshold to obtain the defect focus area; (4) Calculate the area and depth of the defect focus area to obtain defect degree data; (5) Calculate the hazard score based on the defect degree data and screen out defect markers with high hazard degree; (6) The original images corresponding to the high-risk defect marks are extracted and sorted to obtain a set of key defect images.

[0038] Specifically, when normalizing the characteristic values ​​in the pipeline defect feature matrix, the minimum-maximum normalization method is used to standardize each characteristic dimension separately. The purpose of normalization is to unify the characteristic values ​​of different dimensions into the same numerical range and eliminate the dimension effect. For each column in the feature matrix, find the maximum and minimum values ​​of the column respectively, and then map the values ​​to the [0,1] interval according to the linear transformation. The normalized characteristic data maintains the relative relationship of the original data and is comparable. When setting the defect discrimination threshold for the standardized characteristic data, the threshold setting method based on statistical distribution is adopted. A normality test is performed on each characteristic dimension to confirm the data distribution characteristics. Then, based on the 3σ principle, the range of the mean plus or minus three times the standard deviation is used as the normal value domain, and the data points beyond this range are marked as abnormal characteristic points. For each characteristic dimension, the sample mean and standard deviation are calculated, and the upper and lower thresholds are set. When a certain characteristic value exceeds the threshold range of the corresponding dimension, the dimension feature of the data point is marked as an abnormal point.

[0039] When performing spatial clustering analysis on feature points that exceed the discrimination threshold, the density-based spatial clustering algorithm DBSCAN is used. The DBSCAN algorithm divides points with a sufficient number of neighboring points into the same cluster by defining the concepts of core points, direct density reachability, and density reachability. The two key parameters of the algorithm are the neighborhood radius ε and the minimum number of points MinPts. The Euclidean distance between feature points is calculated, and then the number of points in the ε neighborhood of each point is counted, and the points that meet the MinPts requirement are regarded as core points. The relevant core points and their density reachable points are classified into one category through density connectivity, and finally multiple defect focus areas are obtained.

[0040] When calculating the area and depth of the defect area of ​​concern, the following formula is used: ; in: is the defect degree value; M, N are the number of rows and columns of the defect area; is the defect area factor at point (m,n); is the defect depth weight at point (m,n); is the position influence coefficient at point (m,n); γ is the pipe material correction coefficient; Adjust parameters for environmental factors; is the maximum stress concentration factor; is the minimum stress concentration factor.

[0041] When calculating the hazard score based on the defect degree data, a multi-index weighted scoring method is used. The scoring indicator system is determined, including three dimensions: defect degree value, location sensitivity, and development trend. A weight coefficient is set for each dimension to reflect its importance in the overall score. The comprehensive hazard score is obtained by weighted summation. The higher the score, the greater the degree of hazard. In the calculation process, multiple influencing factors such as defect type, pipeline importance, and surrounding environment are considered to ensure the comprehensiveness and accuracy of the scoring results. When extracting and sorting the original images corresponding to the high-hazard defect marks, an index relationship between the hazard score and the original image is established. According to the preset hazard score threshold, defect records above the threshold are screened out. The images corresponding to these high-hazard defects are extracted from the original image database, while retaining the relevant metadata information, including defect code, location information, detection time, etc. The extracted images are stored in a unified format, and a hierarchical directory structure is established for easy management and query.

[0042] For example, the inspection data of a certain section of the pipeline network is analyzed. The original feature matrix contains data of six feature dimensions, such as area, depth, and roundness. After the minimum-maximum normalization process, the standardized feature values ​​are obtained. Taking the area feature as an example, the mean value of 0.5 and the standard deviation of 0.15 are calculated. Based on this, the threshold range [0.05, 0.95] is set, and the points outside this range are marked as abnormal. The DBSCAN algorithm is used to cluster the abnormal points, and the neighborhood radius ε=0.1 and the minimum number of points MinPts=4 are set to obtain multiple defect focus areas. The defect degree value is calculated for each focus area, and the defect degree data is obtained by considering factors such as defect area, depth, and location. The hazard score is calculated by combining factors such as the importance of the pipeline location (main pipe / branch pipe) and the sensitivity of the surrounding environment (residential area / industrial area). The hazard score threshold is set to 0.8, and defect images with scores higher than the threshold are extracted to obtain a set of key defect images, each of which contains defect information and evaluation data.

[0043] In a specific embodiment, the process of executing step S104 may specifically include the following steps: (1) Calculate pixel gradients on the key defect image set to obtain the defect edge contour line; (2) Analyze the shape characteristics of the defect edge contour and extract the defect morphological parameters; (3) Perform numerical segmentation interval statistics on defect morphological parameters to obtain defect type feature vectors; (4) Correlate and match the defect type feature vector with the defect location information to form spatial distribution data; (5) Calculate the defect degree score of the spatial distribution data and generate a defect severity index; (6) Combine the defect severity index with the corresponding morphological characteristics to obtain defect type data.

[0044] Specifically, when calculating pixel gradients for the key defect image set, the following gradient calculation formula is used: ; in: is the gray value at the image pixel (u, v); K, L are the convolution kernel sizes; is the direction weight coefficient; is the edge enhancement factor; is the noise suppression parameter.

[0045] The defect severity score is calculated using: ; Among them: S is the defect severity index; Q, R are the scoring area dimensions; is the defect weight coefficient at position (q, r); is the spatial impact factor at position (q, r); is the morphological characteristic parameter at the position (q, r); ε is the pipe network type correction coefficient; μ is the environmental sensitivity coefficient; is the maximum risk factor; is the minimum risk factor.

[0046] For each image in the key defect image set, gradient calculation is performed. The Sobel operator is used to calculate the gradient in the horizontal and vertical directions, and the gradient components in the two directions are obtained through convolution operations. The gradient amplitude is calculated based on the gradient components, and non-maximum suppression is performed according to the gradient direction information to extract the precise edge contour line. In the edge extraction process, the adaptive threshold method is used to process images under different brightness conditions to ensure the stability of edge detection. The obtained defect edge contour line is subjected to shape feature analysis to extract morphological parameters including contour length, area, perimeter ratio, convexity, eccentricity, etc. The calculation of morphological parameters is based on the contour point set, and the values ​​of various parameters are obtained through matrix operations and geometric calculations. At the same time, the contour is subjected to Fourier descriptor analysis to obtain the frequency domain characteristics of the contour, which is used to characterize the complexity of the defect shape.

[0047] When performing numerical segmentation interval statistics on defect morphological parameters, the data distribution characteristics are used to segment the parameters. For each morphological parameter, the statistical characteristics of its numerical distribution are calculated, including the mean, standard deviation, quantile, etc. According to the physical meaning of the parameter and the numerical distribution characteristics, a reasonable segmentation interval is set to map the continuous parameter value to a discrete interval number. After the morphological parameters of each defect sample are mapped through the interval, a fixed-length feature vector is formed. When associating and matching the defect type feature vector with the defect location information, two types of data need to be processed: one is the feature vector data of the defect, and the other is the spatial location data of the defect. The spatial location data contains information such as the pipe section number, mileage pile number, and pipeline depth. By establishing a corresponding relationship between the feature vector and the location information, spatial distribution data containing complete defect information is formed. The data storage adopts a relational structure to facilitate subsequent spatial analysis and query.

[0048] When calculating the defect severity score for spatially distributed data, the morphological characteristics, location characteristics, and environmental factors of the defects are comprehensively considered according to the above scoring formula. During the scoring process, different weights are assigned to different characteristics to reflect their importance in defect assessment. The scoring results reflect the severity of the defects and provide a basis for subsequent maintenance decisions. When combining the defect severity index with the corresponding morphological characteristics, the data is organized in a vector splicing manner. Each defect record contains the severity index and morphological characteristic information to form a standardized defect type data structure. This data structure facilitates defect classification and retrieval while retaining detailed characteristic information of the defects.

[0049] For example, the deposition defects of a pipe section are analyzed, the gradient of the defect image is calculated, and the 5×5 Sobel operator is used for convolution operation to obtain the gradient map in the horizontal and vertical directions. The edge contour of the defect is extracted by non-maximum suppression and double threshold processing. The shape analysis of the contour is performed to calculate the basic parameters such as contour length and area, and the 8th order frequency domain features are obtained by Fourier descriptor. According to the numerical distribution characteristics of the parameters, the area parameter is divided into 4 intervals, the perimeter ratio is divided into 3 intervals, and other parameters are processed similarly, and finally a 12-dimensional feature vector is obtained. Combined with the location information of the defective pipe section (the pipe section number is M-001 and the mileage pile number is K0+500), the spatial distribution data is formed. Considering that the location is in the main pipeline network and is located in a densely populated area, a higher location weight and environmental sensitivity coefficient are set, and a higher defect degree score is calculated. Finally, the score result is integrated with the morphological feature data to form a defect type data record.

[0050] In a specific embodiment, the process of executing step S105 may specifically include the following steps: (1) Group the defect type data according to the defect morphology to obtain a defect category statistics table; (2) Perform weight assignment calculation on the data in the defect category statistical table to obtain the defect impact factor; (3) Correlate the defect influencing factors with the pipeline operation status data to obtain the defect hazard index; (4) Divide the defect hazard index into graded intervals and generate defect grade determination criteria; (5) Compare and verify the defect grade determination criteria with historical maintenance records to form a defect grade mapping table; (6) Perform data integration and grade matching on the defect grade mapping table to obtain the defect grade assessment result.

[0051] Specifically, when the defect type data is grouped according to the defect morphology, the grouping criteria are established based on the geometric and morphological characteristics of the defects. Defect morphology includes point defects (such as broken points), linear defects (such as cracks), surface defects (such as corrosion areas) and composite defects. For each morphological type, it is further finely grouped according to the characteristic parameters. For example, linear defects are grouped according to directionality (longitudinal, circumferential, oblique) and length range; surface defects are grouped according to area size and shape characteristics (regular, irregular). The grouping results are recorded in the defect category statistics table, which contains information such as defect type, quantity, and distribution characteristics. When the data in the defect category statistics table are weighted, the multi-level analysis method is used to determine the weight coefficient. A weight indicator system is established, including three dimensions: defect size (point, linear, surface), defect location (tube top, tube bottom, tube wall), and defect depth (surface, middle layer, through). Based on expert experience and historical data, a judgment matrix is ​​constructed, consistency test is performed, and the relative importance of each indicator is calculated. The calculated weight coefficient is multiplied by the defect feature data to obtain the defect impact factor.

[0052] When analyzing the correlation between defect influencing factors and pipeline operation status data, pipeline operation parameters are introduced, including pipeline service life, operating pressure, flow load, etc. The correlation analysis method is used to calculate the correlation between defect influencing factors and various operation parameters. The defect hazard index is obtained by weighted summation, which reflects the degree of influence of defects on pipeline operation safety. At the same time, external factors such as the importance of the area where the pipeline is located and the sensitivity of the surrounding environment are considered to correct the hazard index. When dividing the defect hazard index into graded intervals, the cluster analysis method is used to determine the grade boundaries. The K-means algorithm is used to cluster the hazard index data, and 3-5 grade intervals are initially set. The optimal number of classifications is determined by minimizing the intra-class variance and maximizing the inter-class variance. The corresponding defect level is set for each grade interval, and the defect level judgment standard is established. The division results need to take into account the actual engineering needs to ensure the practicality and operability of the classification results.

[0053] When comparing and verifying the defect grade determination criteria with historical maintenance records, extract the defect grade assessment and processing records in the historical maintenance database. For each defect grade, count the distribution of historical maintenance measures, and analyze the consistency between the grade determination results and the actual maintenance needs. Through comparative analysis, establish a mapping relationship between defect grades and maintenance measures to form a standardized defect grading mapping table. When performing data integration and grade matching on the defect grading mapping table, match the newly identified defect data with the mapping table. According to the characteristic parameters and hazard index of the defect, find the closest match in the mapping table to determine the defect grade. For special or new defects, use the similarity calculation method to find the closest historical case for grade determination. The final output is an evaluation result containing complete information such as defect type, grade, and recommended maintenance measures.

[0054] For example, a section of a drainage network was inspected and multiple defects were found. The defects were morphologically classified and two point damages, one longitudinal crack, and one surface corrosion area were identified. Through geometric feature analysis, the length of the longitudinal crack is 0.3 times the circumference of the pipeline, and the corrosion area covers 0.2 times the surface area of ​​the pipe wall. The weights were calculated using the multi-level analysis method, with the position weight of crack defects being 0.4, the depth weight being 0.3, and the size weight being 0.3. Combined with the defect characteristic values, the impact factor of the crack was calculated. The operation life of the pipe section is 15 years, and the average daily flow is 0.8 times the design flow. The operation status correction coefficient was obtained through correlation analysis. The defect hazard index was obtained through comprehensive calculation, and the hazard index was divided into four level intervals through K-means clustering. Compared with historical maintenance records, defects with similar hazard indexes are usually repaired by excavation. The final evaluation results show that the crack defect belongs to the second level and needs to be repaired as a priority.

[0055] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) Compare the defect level assessment results with the maintenance technology solution library to form a maintenance measure correspondence table; (2) Cost accounting is performed on the maintenance plans in the maintenance measures correspondence table to obtain a maintenance resource requirement list; (3) Prioritize the maintenance resource requirement list according to the severity of the defects and generate a maintenance schedule; (4) Conduct construction condition constraint analysis on the maintenance schedule and form a feasibility assessment report; (5) Match the feasibility assessment report with the existing maintenance resources to obtain a maintenance scheduling plan; (6) Comprehensively optimize and integrate the maintenance scheduling plan to obtain the drainage network maintenance decision plan.

[0056] Specifically, when comparing the defect level assessment results with the maintenance technology solution library, the maintenance technology solution library contains standard treatment measures for various defects, such as excavation repair, in-situ solidification repair, local repair, etc. For each defect level assessment result, the corresponding maintenance technology solution is retrieved and matched to form a maintenance measure correspondence table. The corresponding table records information such as defect type, level, location, recommended maintenance method, technical parameters, etc. For complex defects, it is necessary to combine multiple maintenance measures to form a comprehensive solution. When costing the maintenance solutions in the maintenance measure correspondence table, both direct costs and indirect costs are considered. Direct costs include material costs, labor costs, equipment costs, etc. When calculating specifically, factors such as engineering volume, material unit price, labor unit price, and equipment unit cost need to be considered. Indirect costs include traffic diversion fees, environmental protection fees, temporary facility fees, etc. For each maintenance plan, a detailed cost structure is listed and a maintenance resource demand list is compiled. When the maintenance resource demand list is prioritized according to the severity of the defect, a multi-factor comprehensive scoring method is adopted. The scoring indicators include defect level, pipeline importance, impact range, construction difficulty, etc. Assign weight coefficients to each indicator and obtain a comprehensive score through weighted calculation. Sort by scores and arrange maintenance schedules reasonably. For maintenance projects with dependencies, the logical order of construction needs to be considered. When analyzing the construction condition constraints for maintenance schedules, multiple constraints need to be considered. Construction environment constraints include underground pipeline distribution, traffic flow, surrounding buildings, etc.; climate condition constraints include rainfall, temperature, etc.; resource constraints include equipment, personnel, materials, etc. By systematically analyzing the impact of various constraints on maintenance projects, a feasibility assessment report containing construction risks, technical difficulties, resource requirements, etc. is formed. When matching the feasibility assessment report with existing maintenance resources, an inventory of existing maintenance resources is taken, including construction teams, professional equipment, and material reserves. Determine resource gaps through comparative analysis, reasonably allocate existing resources, and supplement deficiencies through external introduction when necessary. Prepare a maintenance scheduling plan that includes resource allocation plans, construction schedules, quality control measures, etc.

[0057] When optimizing and integrating the maintenance scheduling plan, the construction network diagram and critical path analysis method are used. Each maintenance project is organized into a network diagram according to the logical relationship, the critical path is calculated, and the construction process is optimized. A comprehensive balance is made on time schedule, resource allocation, cost control and other aspects to form the final drainage network maintenance decision plan.

[0058] For example, three defects were detected in a certain trunk section, including two secondary cracks and one tertiary deposit. After comparison with the maintenance technology solution library, the cracks were repaired by in-situ curing and the deposits were repaired by high-pressure cleaning. Cost accounting shows that in-situ curing repairs require 30,000 yuan in material costs, 15,000 yuan in labor costs, and 20,000 yuan in equipment costs for each location, and high-pressure cleaning requires 8,000 yuan in equipment costs and 5,000 yuan in labor costs. Through multi-factor scoring, the crack defect scores were 85 and 82 points respectively, and the sediment defect score was 76 points, and the construction sequence was arranged accordingly. Analysis of the construction conditions found that the pipe section was located on a major traffic artery, required nighttime construction, and was affected by the rainy season. According to the existing resource situation, a curing and repair team and a dredging team were arranged to prepare a phased implementation plan, which was optimized to form a complete maintenance decision plan including construction organization, quality control, and emergency plans.

[0059] The above describes the method for automatically identifying and grading the functional defects of urban drainage pipe networks in the embodiment of the present application. The following describes the system for automatically identifying and grading the functional defects of urban drainage pipe networks in the embodiment of the present application. Figure 3 In the embodiment of the present application, an embodiment of the automatic identification and maintenance classification system of functional defects of urban drainage pipe networks includes: The acquisition module 201 is used to obtain the drainage network inspection video through the pipeline inspection equipment, and encode and extract the inspection video according to the six-dimensional information to obtain the pipeline network defect image data, wherein the six-dimensional information includes city, region, year, defect name, level and serial number; An extraction module 202 is used to perform image decoding and feature extraction processing on the pipeline network defect image data to obtain a pipeline defect feature matrix; The screening module 203 is used to perform screening processing on key defects according to the pipeline defect feature matrix to obtain a set of key defect pictures, including: setting a discrimination threshold after normalizing the values ​​in the pipeline defect feature matrix, extracting feature points exceeding the threshold and performing spatial clustering analysis to obtain a defect focus area; calculating a hazard score according to the area and depth of the defect focus area, screening out defects with high hazard levels and extracting corresponding original pictures to obtain a set of key defect pictures; The quantization module 204 is used to perform feature refinement and parameter quantification processing on the key defect image set to obtain defect type data; The classification module 205 is used to perform grade evaluation and classification processing on the defect type data to obtain a defect grade evaluation result; The sorting module 206 is used to match and prioritize maintenance plans according to the defect level assessment results to obtain a drainage network maintenance decision plan.

[0060] Through the collaborative cooperation of the above-mentioned components, the drainage network inspection video is obtained through the pipeline inspection equipment, and the encoding, extraction and processing are carried out according to the six-dimensional information. A standardized defect data collection and management mechanism is established, and the standardized storage and efficient retrieval of the inspection data are realized, laying a data foundation for subsequent intelligent analysis; the feature matrix of pipeline defects is obtained by image decoding and feature extraction processing of the pipeline network defect picture data, making the defect feature extraction process more objective and comprehensive, avoiding the subjectivity and inconsistency of manual interpretation; the pipeline defect feature matrix is ​​screened for key defects, improving the recognition and attention of important defects, and effectively preventing important defects from being omitted or ignored; the key defect picture set is subjected to feature refinement and parameter quantification processing, realizing the accurate quantification of defect features, and providing reliable data support for defect assessment; the defect type data is graded and classified, and a scientific defect grading system is established, making the defect grade determination more accurate and objective; the maintenance plan is matched and prioritized according to the defect grade assessment results, realizing the scientific and automated maintenance decision-making, improving the rationality and implementation effect of the maintenance plan, and greatly improving the work efficiency and accuracy of drainage network defect identification and maintenance management.

[0061] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for automatic identification and maintenance classification of functional defects in urban drainage networks, characterized in that: The method for automatically identifying and grading functional defects of urban drainage pipe networks includes: Obtain drainage network inspection video through pipeline inspection equipment, encode and extract the inspection video according to six-dimensional information, and obtain network defect image data, wherein the six-dimensional information includes city, region, year, defect name, grade and serial number; Performing image decoding and feature extraction processing on the pipeline network defect image data to obtain a pipeline defect feature matrix; Performing key defect screening according to the pipeline defect feature matrix to obtain a key defect picture set, including: normalizing the values ​​in the pipeline defect feature matrix and setting a discrimination threshold, extracting super-threshold feature points and performing spatial clustering analysis to obtain a defect focus area; calculating a hazard score according to the area and depth of the defect focus area, screening out defects with high hazard levels and extracting corresponding original pictures to obtain a key defect picture set; Perform feature refinement and parameter quantification processing on the key defect image set to obtain defect type data; Performing grade assessment and classification processing on the defect type data to obtain a defect grade assessment result; Maintenance scheme matching and priority sorting are performed according to the defect level assessment results to obtain a drainage network maintenance decision plan.

2. The method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks according to claim 1 is characterized in that: The drainage network inspection video is obtained by the pipeline inspection equipment, and the inspection video is encoded and extracted according to the six-dimensional information to obtain the network defect image data, wherein the six-dimensional information includes the city, region, year, defect name, level and serial number, including: Capturing video of the drainage network through pipeline CCTV detection equipment, and extracting frames of the drainage network detection video according to a preset sampling frequency; Perform noise reduction and illumination compensation preprocessing on the frame image sequence of the detection video to remove image noise and the influence of uneven illumination; Extracting key frame images from the preprocessed frame image sequence, and establishing a spatiotemporal index for the key frame images; The key frame images are coded and identified according to the city, region, year, defect name, grade and serial number in the six-dimensional information; Associating and matching the coding identifier with the corresponding key frame image to generate image data with a unique identifier; The image data with unique identifiers are batch sorted and format converted to form the pipeline network defect image data.

3. The method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks according to claim 1 is characterized in that: The image decoding and feature extraction processing of the pipeline network defect image data is performed to obtain a pipeline defect feature matrix, including: Performing pixel distribution statistical analysis on the pipeline network defect image data to obtain image grayscale distribution data; The image grayscale distribution data is processed by threshold segmentation to obtain a binary image of the defect area; Performing edge detection and contour extraction on the defect area binary image to obtain a defect boundary point set; Calculate the geometric features of the defect boundary point set to extract defect area, perimeter, and roundness measurement values; Performing texture feature analysis on the defect area to extract energy, contrast, and correlation parameters of the gray-level co-occurrence matrix; The geometric features and texture features are data-fused to obtain the pipeline defect feature matrix.

4. The method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks according to claim 1 is characterized in that: The key defect screening process is performed according to the pipeline defect feature matrix to obtain a key defect picture set, including: Normalizing the characteristic values ​​in the pipeline defect characteristic matrix to obtain standardized characteristic data; Setting a defect discrimination threshold for the standardized feature data, and extracting feature points exceeding the discrimination threshold; Performing spatial cluster analysis on the feature points exceeding the discrimination threshold to obtain defect focus areas; Calculating the area and depth of the defect concern area to obtain defect degree data; Calculate the hazard score based on the defect degree data and select defect marks with high hazard degree; The original images corresponding to the high-hazard degree defect marks are extracted and sorted to obtain the key defect image set.

5. The method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks according to claim 1 is characterized in that: The key defect picture set is subjected to feature refinement and parameter quantification processing to obtain defect type data, including: Calculating pixel gradients on the key defect image set to obtain defect edge contours; Performing shape feature analysis on the defect edge contour line to extract defect morphological parameters; Performing numerical segmented interval statistics on the defect morphological parameters to obtain a defect type feature vector; Associating and matching the defect type feature vector with the defect location information to form spatial distribution data; Performing defect degree scoring calculation on the spatial distribution data to generate a defect severity index; The defect severity index is combined with the corresponding morphological feature to obtain the defect type data.

6. The method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks according to claim 1 is characterized in that: The step of performing grade assessment and classification processing on the defect type data to obtain a defect grade assessment result includes: The defect type data are grouped according to the defect morphology to obtain a defect category statistics table; Perform weight assignment calculation on the data in the defect category statistical table to obtain defect impact factors; Correlation analysis is performed on the defect influencing factor and pipeline operation status data to obtain a defect hazard index; Dividing the defect hazard index into graded intervals to generate defect grade determination criteria; Compare and verify the defect grade determination standard with historical maintenance records to form a defect grade mapping table; Data integration and grade matching are performed on the defect grade mapping table to obtain the defect grade evaluation result.

7. The method for automatic identification and maintenance classification of functional defects of urban drainage pipe networks according to claim 1 is characterized in that: The maintenance scheme matching and priority sorting processing are performed according to the defect level assessment results to obtain a drainage network maintenance decision plan, including: Compare the defect level assessment results with the maintenance technology solution library to form a maintenance measure correspondence table; Cost accounting is performed on the maintenance plans in the maintenance measure correspondence table to obtain a maintenance resource requirement list; Prioritize the maintenance resource requirement list according to the severity of defects and generate a maintenance schedule; Conduct construction condition constraint analysis on the maintenance schedule and form a feasibility assessment report; Matching and analyzing the feasibility assessment report with existing maintenance resources to obtain a maintenance scheduling plan; The maintenance scheduling plan is comprehensively optimized and integrated to obtain the drainage network maintenance decision plan.

8. A system for automatically identifying and grading the functional defects of an urban drainage network, used to implement the method for automatically identifying and grading the functional defects of an urban drainage network as claimed in any one of claims 1 to 7, characterized in that: The automatic identification and maintenance classification system for functional defects of urban drainage pipe networks includes: The acquisition module is used to obtain the drainage network inspection video through the pipeline inspection equipment, encode and extract the inspection video according to the six-dimensional information, and obtain the pipeline network defect image data, wherein the six-dimensional information includes city, region, year, defect name, level and serial number; An extraction module is used to perform image decoding and feature extraction processing on the pipeline network defect image data to obtain a pipeline defect feature matrix; A screening module is used to screen key defects according to the pipeline defect feature matrix to obtain a key defect picture set, including: setting a discrimination threshold after normalizing the values ​​in the pipeline defect feature matrix, extracting super-threshold feature points and performing spatial clustering analysis to obtain a defect focus area; calculating a hazard score according to the area and depth of the defect focus area, screening out defects with high hazard levels and extracting corresponding original pictures to obtain a key defect picture set; A quantification module is used to perform feature refinement and parameter quantification processing on the key defect image set to obtain defect type data; A classification module, used to perform grade assessment and classification processing on the defect type data to obtain a defect grade assessment result; The sorting module is used to match and prioritize maintenance plans according to the defect level assessment results to obtain a drainage network maintenance decision plan.

Images