An intelligent generation method and system for urban sewer network defect repair suggestions

Abstract

The application discloses a kind of urban drainage pipe network defect repair suggestion intelligent generation method and system, it is related to numerical, image, text and other multi-modal processing technical field, including: acquisition drainage pipeline CCTV video dataset utilizes defect category detection and segmentation model, obtains drainage pipe network defect detection segmentation result;Based on the drainage pipe network defect detection segmentation result combines preset instruction template, obtains drainage pipe network defect repair suggestion database;According to the drainage pipe network defect detection segmentation result and the drainage pipe network defect repair suggestion database, obtain urban drainage pipe network defect repair suggestion intelligent generation result.The application solves the core pain points of long time sequence video noise accumulation, defect identity drift, detection and repair process disconnection, lack of spatial carrier of pipe network data in the prior art, realizes the integration solution from pipe network data acquisition, AI defect identification to intelligent generation professional repair scheme.

Description

Technical Field

[0001] This invention relates to the field of multimodal processing technology, including numerical, image, and text processing, and particularly to an intelligent method and system for generating suggestions for repairing defects in urban drainage pipe networks. Background Technology

[0002] Urban drainage networks are essential infrastructure for ensuring the normal operation and healthy development of cities, including flood control, disaster mitigation, waste disposal, and sewage treatment. They maintain the overall operation of the city's underground and above-ground spaces and are a crucial tool for risk prevention and improving people's livelihoods; hence, they are often referred to as the "lifeline project" of a city. With the deepening of urbanization, road collapses and urban flooding have become increasingly frequent, exposing numerous deficiencies in urban drainage networks and hindering the progress of smart city construction.

[0003] Artificial intelligence (AI) is an interdisciplinary field integrating computer science, mathematics, linguistics, and other disciplines. Its core goal is to enable computers to possess intelligent behaviors similar to humans, such as perception, learning, reasoning, and decision-making. Convolutional neural networks (CNNs) automatically extract local features through local connectivity, weight sharing, and translation invariance, thereby improving the model's generalization ability and computational efficiency. Common types include one-dimensional and two-dimensional convolutions. One-dimensional convolutions extract local features by sliding the convolution kernel along the time dimension, have relatively fewer parameters, and can capture and process temporal dependencies and dynamic changes. Two-dimensional convolutions extract local features by sliding the convolution kernel along the spatial dimension, have relatively more parameters, and are good at understanding and processing data with spatial structure. Transformer networks process the entire sequence in parallel through a self-attention mechanism, are good at capturing long-distance dependencies, and have powerful contextual semantic understanding and generation capabilities. Thanks to breakthroughs in large language models and tools, intelligent agents can perform autonomous analysis and decision-making through multi-turn dialogue and memory techniques, and are developing towards more natural interactions and broader collaborations.

[0004] However, the internal environment of urban drainage pipes is complex and variable, with diverse shapes and sizes of defects that persist for extended periods in videos. These factors significantly hinder the detection and segmentation of defect categories. Furthermore, current methods for detecting and segmenting drainage network defects only reach the "problem discovery" stage, failing to connect with the repair and assessment phase. This leads to a break in the closed-loop management of the network, weakening its identification value and restricting the precision and long-term effectiveness of network maintenance. Therefore, there is an urgent need for an intelligent method and system for generating repair suggestions for urban drainage network defects to address the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to propose an intelligent generation method and system for urban drainage network defect repair suggestions, in order to solve problems such as noise accumulation in long-term video, identity drift, disconnect between detection segmentation and repair, and lack of spatial carrier for network data.

[0006] Firstly, to achieve the above objectives, the present invention provides an intelligent method for generating suggestions for repairing defects in urban drainage pipe networks, comprising:

[0007] S1. Collect CCTV video dataset of drainage pipelines and use the defect category detection and segmentation model to obtain the defect detection and segmentation results of drainage pipeline network;

[0008] S2. Based on the drainage network defect detection and segmentation results and the preset instruction template, obtain a drainage network defect repair suggestion database;

[0009] S3. Based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, obtain the intelligent generation result of urban drainage network defect repair suggestions.

[0010] Optionally, S1, collect CCTV video dataset of drainage pipes and use a defect category detection and segmentation model to obtain defect detection and segmentation results of the drainage pipe network, including:

[0011] Data preprocessing of CCTV video sets of drainage pipes is performed using the defect categories of drainage pipe networks to obtain target video segments of drainage pipes;

[0012] The target video segments of the drainage pipeline are labeled according to the frame-by-frame annotation density to establish a target sample dataset of the drainage pipeline;

[0013] Based on the target sample dataset of the drainage pipe, a deep learning algorithm is used to train and construct a defect category detection and segmentation model. The deep learning algorithm includes a segmentation network, a tracking network, a correction network, and a decoding network.

[0014] The defect category detection and segmentation model is used to detect and segment the CCTV video set of the drainage pipe to obtain the defect category and the area ratio of the drainage pipe defect.

[0015] Based on the proportion of the defect area in the drainage pipe, the defect level of the drainage pipe is determined, and by combining the defect category of the drainage pipe with the proportion of the defect area in the drainage pipe, the defect detection and segmentation results of the drainage pipe network are obtained.

[0016] Optionally, a defect category detection and segmentation model is constructed by training a deep learning algorithm based on the target sample dataset of the drainage pipe, including:

[0017] Input the target sample dataset of the drainage pipe into the segmentation network to obtain the defect feature query vector output by the segmentation network;

[0018] Based on the defect feature query vector output by the segmentation network, the tracking network is used for dynamic tracking to obtain the defect tracking result output by the tracking network;

[0019] The defect tracking results output by the tracking network are input into the correction network for defect calibration, and the defect features optimized by the correction network are obtained.

[0020] Based on the defect features optimized by the modified network, the decoding network is used to decode and obtain defect repair suggestions output by the decoding network.

[0021] Based on the defect repair suggestions output by the decoding network, the deep learning algorithm is used to iteratively train the target sample dataset of the drainage pipe to obtain a defect category detection and segmentation model.

[0022] Optionally, S2, based on the drainage network defect detection and segmentation results combined with a preset instruction template, obtain a drainage network defect repair suggestion database, including:

[0023] The defect detection and segmentation results of the drainage pipe network are structured and organized to obtain the core defect information.

[0024] Based on the structured and organized core defect information combined with the preset instruction template, standardized instruction text is obtained;

[0025] Based on the standardized instruction text, the intelligent agent performs autonomous planning and task decomposition to obtain preliminary repair process suggestions.

[0026] The initial repair process suggestions are verified and optimized based on the drainage network defect detection and segmentation results to obtain a drainage network defect repair suggestion database.

[0027] Optionally, S3, based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, obtain the intelligent generation result of urban drainage network defect repair suggestions, including:

[0028] Using the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, basic archive information of drainage pipelines is obtained;

[0029] Based on the basic archive information of the drainage pipeline, CCTV video data of the current drainage pipeline is created to obtain drainage pipeline data;

[0030] The drainage pipeline data is processed and analyzed to obtain related data from the processing and analysis;

[0031] Based on the associated data processed and analyzed, the defect category detection and segmentation model is used to obtain the current drainage network defect detection and segmentation results.

[0032] Based on the current drainage network defect detection and segmentation results and the drainage pipeline basic file information, a defect assessment is performed to obtain intelligent generation results of urban drainage network defect repair suggestions.

[0033] Optionally, based on the basic archive information of the drainage pipeline, CCTV video data of the current drainage pipeline is created to obtain drainage pipeline data, including:

[0034] Based on the CCTV video data of the current drainage pipeline, obtain the current drainage pipeline information;

[0035] The current drainage pipeline information is compared and analyzed with the basic drainage pipeline file information to obtain the comparison result of the current drainage pipeline information;

[0036] Based on the comparison results of the current drainage pipeline information, determine whether the current drainage pipeline is a newly added pipeline. If it is, execute the first operation; otherwise, execute the second operation.

[0037] The first operation is: updating the CCTV video set of the drainage pipeline based on the current CCTV video data of the drainage pipeline, and performing the third operation;

[0038] The second operation is: to retrieve the drainage pipeline basic file information based on the current drainage pipeline information to obtain the drainage pipeline information;

[0039] The third operation is to preprocess the CCTV video set of drainage pipes using the drainage pipe network defect category to obtain the target video segment of the drainage pipes.

[0040] Secondly, to achieve the above objectives, the present invention provides an intelligent generation system for suggestions on repairing defects in urban drainage pipe networks, comprising:

[0041] Defect detection and segmentation module, repair suggestion generation module, and data processing and analysis module;

[0042] The defect detection and segmentation module is used to collect CCTV video datasets of drainage pipelines and use a defect category detection and segmentation model to obtain defect detection and segmentation results of the drainage pipeline network.

[0043] The repair suggestion generation module is used to obtain a database of repair suggestions for drainage network defects based on the drainage network defect detection and segmentation results and a preset instruction template.

[0044] The data processing and analysis module is used to obtain intelligent generation results of urban drainage network defect repair suggestions based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database.

[0045] Optionally, the data processing and analysis module includes a data loading submodule, a data creation submodule, a data processing and analysis submodule, a data detection and segmentation submodule, and a data defect assessment submodule;

[0046] The data loading submodule is used to obtain basic archive information of drainage pipelines by utilizing the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database.

[0047] The data creation submodule is used to create CCTV video data of the current drainage pipeline based on the basic archive information of the drainage pipeline, and to obtain drainage pipeline data;

[0048] The data processing and analysis submodule is used to process and analyze the drainage pipeline data and obtain related data from the processing and analysis.

[0049] The data detection and segmentation submodule is used to obtain the current drainage network defect detection and segmentation results based on the processed and analyzed associated data and the defect category detection and segmentation model.

[0050] The data defect assessment submodule is used to assess defects based on the current drainage network defect detection and segmentation results combined with the basic archive information of the drainage pipeline, and to obtain intelligent generation results of urban drainage network defect repair suggestions.

[0051] Thirdly, to achieve the above objectives, the present invention provides an electronic device, comprising: one or more processors; and a storage device having stored one or more programs thereon, which, when executed by the one or more processors, cause the one or more processors to implement the method described in any implementation of the first aspect.

[0052] Fourthly, to achieve the above objectives, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any implementation of the first aspect.

[0053] Compared with the closest existing technology, the present invention has the following advantages:

[0054] This invention addresses the core pain points of existing technologies, such as long-term video noise accumulation, defect identity drift, disconnect between detection and repair processes, and lack of spatial carrier for pipeline network data. It constructs a fully intelligent closed loop of "perception-planning-execution-evaluation-iteration", realizing an integrated solution from pipeline network data acquisition and AI defect identification to intelligent generation of professional repair solutions.

[0055] At the methodological level, on the one hand, a decoupling strategy is adopted to construct a multi-module collaborative deep learning model, which effectively blocks the propagation path of errors in the long-term pipeline network video detection and segmentation chain, solves the problems of noise accumulation and identity drift, and significantly improves the accuracy and robustness of defect identification. On the other hand, relying on a large language model-driven intelligent agent, it accurately understands user needs and autonomously integrates multi-source heterogeneous data such as pipeline network defect detection technical standards, evaluation reports, and historical cases. Combined with preset instruction templates, it achieves seamless connection between defect data and professional repair suggestions, breaks down the information barriers between detection and repair links, and outputs standardized and traceable professional repair suggestions.

[0056] At the system level, the GIS-based visualization management platform serves as a unified spatial carrier for pipeline network data. It spatially associates attribute information, video data, defect results, and repair solutions, enabling seamless data access to smart drainage pipeline network platforms across various regions. This addresses the problem of traditional pipeline network data lacking a unified spatial platform and being difficult to integrate into the smart drainage system. Users can complete the entire process from data loading and defect identification to repair suggestion generation through the software interface, significantly lowering the application threshold and achieving more efficient and accurate pipeline defect management. This provides solid integrated technical support for the refined, intelligent operation and maintenance and scientific decision-making of urban drainage pipeline networks. Attached Figure Description

[0057] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0058] Figure 1 This is a flowchart of an intelligent method for generating suggestions for repairing defects in urban drainage pipe networks, according to an embodiment of the present invention.

[0059] Figure 2 This is an overall flowchart proposed in an embodiment of the present invention;

[0060] Figure 3 This is a network structure diagram proposed in an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of the structure of an intelligent system for generating suggestions for repairing defects in urban drainage pipe networks, as proposed in an embodiment of the present invention.

[0062] Figure 5 This is a functional architecture diagram of the urban drainage network data processing and analysis software proposed in an embodiment of the present invention;

[0063] Figure 6This is a schematic diagram of the structure of the electronic device proposed in an embodiment of the present invention. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0065] The terminology used in the embodiments section of this invention is for the purpose of explaining specific embodiments of the invention only, and is not intended to limit the invention.

[0066] This invention targets video data of drainage pipe network defects. By combining convolutional neural networks and Transformer networks, it assigns a unique identifier to the same defect throughout the entire video sequence, ensuring cross-frame consistency of defect category detection and segmentation results. Based on these results, the agent autonomously plans and decomposes tasks, and through wide-area information retrieval, outputs structured suggestions for pipe network defect repair.

[0067] like Figure 1 As shown, this embodiment of the invention provides a method for intelligently generating suggestions for repairing defects in urban drainage pipe networks, including:

[0068] S1. Collect CCTV video dataset of drainage pipelines and use the defect category detection and segmentation model to obtain the defect detection and segmentation results of drainage pipeline network;

[0069] This step involves collecting CCTV video datasets of drainage pipes. First, the videos are preprocessed to extract 15-second target segments containing defects. Then, defect labels, IDs, and mask annotations are completed using a frame-by-frame annotation method to construct a sample dataset. Subsequently, a deep learning algorithm that integrates two-dimensional convolutional and Transformer networks is used to train the model. The trained defect category detection and segmentation model is then used to detect and segment the CCTV video dataset of drainage pipes, outputting the defect category and defect area ratio. Based on the area ratio, the defect level is determined, ultimately forming a complete result for the detection and segmentation of defects in the drainage pipe network.

[0070] S2. Based on the drainage network defect detection and segmentation results and the preset instruction template, obtain a drainage network defect repair suggestion database;

[0071] This step, based on the results of drainage network defect detection and segmentation, first structures and organizes core information such as defect type, level, pipe diameter, material, and burial depth. Then, the organized data is filled into a pre-set instruction template containing complete elements such as task objectives, pipeline information, constraints, technical specifications, and output requirements, generating standardized instruction text. Next, the instruction text is input into an intelligent agent centered on a large language model. The agent then performs autonomous planning, problem discovery, wide-area information retrieval, and resource integration, initially outputting repair process suggestions adapted to site conditions. Finally, the initial suggestions are verified for compliance and construction feasibility, and optimized and adjusted based on the defect detection and segmentation results, forming professional, implementable, and industry-standard drainage network defect repair suggestions. This leads to the construction of a drainage network defect repair suggestion database covering all types of defects and all applicable scenarios, providing data support for the intelligent generation, rapid retrieval, and continuous optimization of subsequent drainage network defect repair solutions.

[0072] S3. Based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, obtain the intelligent generation result of urban drainage network defect repair suggestions;

[0073] Integrating defect detection and segmentation results with defect repair suggestions, the system first extracts basic archive information such as pipe diameter, material, service life, and road section of the target drainage pipeline. Then, based on the archive information, it creates or updates the drainage pipeline data corresponding to the current CCTV video. Through spatial processing analysis such as coordinate transformation and topology analysis, it obtains related data. Subsequently, it calls the defect category detection and segmentation model again to obtain the latest defect detection and segmentation results. Combining the pipeline's basic archives, it completes defect assessment and spatial integration. Finally, it outputs a complete result of intelligently generating urban drainage network defect repair suggestions, which includes defect information, repair plans, priorities, and acceptance indicators. This result can be visualized and recorded in the attribute table of the GIS system.

[0074] like Figure 2As shown, the entire technical path from data acquisition and model training to intelligent generation of repair suggestions is presented: First, basic information such as pipe diameter and material is extracted based on the urban drainage network attribute table and associated with CCTV videos of the network. At the same time, a sample dataset is constructed by manually selecting and labeling video clips of pipeline defects, which is used to train a defect detection and segmentation model that integrates two-dimensional convolutional and Transformer networks. Then, the CCTV videos of the network are input into the trained model, which automatically identifies and outputs network defect information containing defect categories and levels. Next, the defect information is combined with user instructions and input into an intelligent agent composed of modules for autonomous planning, problem discovery, resource integration, and conclusion generation. After information structuring, instruction generation, and scheme verification and optimization, network defect repair suggestions adapted to the scenario are formed. The entire process realizes a closed loop from basic network data, video acquisition, AI defect identification to intelligent generation of professional repair solutions, providing complete technical support for subsequent GIS visualization and management applications.

[0075] As one possible implementation, in the above embodiments, step S1 may specifically include the following steps:

[0076] S1-1. Preprocess the CCTV video set of drainage pipes using the defect categories of drainage pipe network to obtain the target video segments of drainage pipes.

[0077] Based on the drainage pipe network defect categories shown in Table 1, the defect videos in the collected drainage pipe CCTV video dataset were preprocessed, that is, the defect videos in the collected drainage pipe CCTV video dataset were cropped. In order to include the complex internal environment of the pipes as comprehensively as possible, and to reduce the amount of annotation work, 15-second video clips containing drainage pipe network defect categories were selected.

[0078] Table 1

[0079] Serial Number Category Serial Number Category 1 Rupture 9 Foreign object penetration 2 Deformation 10 Leakage 3 Corrosion 11 Deposition 4 Misalignment 12 Fouling 5 Undulation 13 Obstruction 6 Dislocation 14 Residual wall, dam root 7 Interface material shedding 15 Tree root 8 Branch pipe hidden joint 16 Dross

[0080] S1-2. Label the target video segments of the drainage pipe according to the frame-by-frame labeling density, and establish the target sample dataset of the drainage pipe;

[0081] High-quality annotation was performed on 15-second video clips containing drainage pipe network defect categories according to the frame-by-frame annotation density. Each category's ID, label, and mask were annotated, and strict quality control was implemented to create a sample dataset containing defect categories. This provides high-quality annotation support for training defect category detection and segmentation models. The frame-by-frame annotation density refers to the density of frame annotation in a video sequence at fixed intervals, obtained by extracting video frames at preset intervals (e.g., every 1 frame, every 2 frames) and completing the annotation.

[0082] S1-3. Based on the target sample dataset of the drainage pipe, a deep learning algorithm is used to train and construct a defect category detection and segmentation model;

[0083] Based on a sample dataset containing defect categories, a defect category detection and segmentation model was constructed through iterative training using a deep learning algorithm that integrates convolutional neural networks and Transformer networks. The deep learning algorithm consists of a segmentation network, a tracking network, a correction network, and a decoding network. The segmentation network uses 2D convolution combined with Transformer to extract multi-scale defect features and potential regions within a single frame. The tracking network models cross-frame tracking as a denoising task to achieve initial correspondence between defect instances. The correction network uses 1D convolution combined with temporal self-attention to calibrate defect features and avoid identity drift. The decoding network outputs defect boxes and masks through the detection head and segmentation head, respectively. The model was trained using NVIDIA RTX A6000 computer hardware resources and the PyTorch deep learning framework, with a 4:1 dataset ratio, a batch size of 8, 50 training epochs, a learning rate of 0.0001, weight decay of 0.05, and the parameters of the AdamW optimizer, resulting in a stable and usable model.

[0084] S1-4. Use the defect category detection and segmentation model to detect and segment the CCTV video set of the drainage pipe to obtain the defect category and the area ratio of the drainage pipe defect.

[0085] The CCTV video set of drainage pipes is input into the trained defect category detection and segmentation model. Through the model's automated inference processing, the internal defects of the pipes are accurately identified and analyzed at the pixel level. The model outputs the defect category corresponding to the drainage pipe in the video and calculates the defect area ratio of each defect within the detection range, providing accurate and quantitative data support for subsequent defect level determination.

[0086] S1-5. Determine the defect level of the drainage pipe based on the proportion of the defect area of ​​the drainage pipe, and obtain the drainage network defect detection and segmentation results by combining the defect category of the drainage pipe with the proportion of the defect area of ​​the drainage pipe.

[0087] Based on the defect area ratio output by the model, the defects are divided into four levels: 1 to 4. Then, the three key pieces of information, namely defect category, defect area ratio, and defect level, are integrated to form a complete drainage network defect detection and segmentation result that includes qualitative identification, quantitative calculation, and level assessment. This provides standardized and structured input data for the generation of subsequent repair suggestions.

[0088] In summary, steps S1-1 to S1-5 sequentially complete the preprocessing of CCTV video of drainage pipelines, the construction of sample datasets through frame-by-frame annotation, the training of a defect category detection and segmentation model based on a deep learning algorithm that integrates convolutional neural networks and Transformers, the acquisition of defect category and area ratio through model inference, the determination of defect levels 1-4 based on the ratio, and finally the integration to form a complete defect detection and segmentation result for drainage pipeline network. The overall process can effectively cover 16 types of pipeline network defects, suppress the accumulation of long-term video noise and identity drift problems, significantly improve the accuracy and robustness of defect detection, segmentation, tracking and classification, and provide stable and reliable standardized data support for the intelligent generation of subsequent repair suggestions.

[0089] like Figure 3 As shown, as a possible implementation, in the above embodiments, steps S1-3 may specifically include the following steps:

[0090] S1-3-1. Input the target sample dataset of the drainage pipe into the segmentation network to obtain the defect feature query vector output by the segmentation network;

[0091] The segmentation network (Seg), serving as the front-end perceptual unit of the model, first utilizes a 2D convolutional neural network to extract multi-scale feature maps rich in contextual information from a single-frame CCTV image of a drainage pipe. Then, a Transformer network introduces a fixed set of learnable query vectors. These query vectors are continuously refined through the interaction of multiple Transformer layers and establish connections with image features via a cross-attention mechanism, thereby selectively capturing potential defect regions in the image. For each input image Efi frame, after processing by the segmentation network, X query vectors for that frame are obtained, with the following relationship:

[0092] Q Seg =Seg(Efi)

[0093] Among them, Q Seg Let Seg(⋅) be the defect feature query vector output by the segmentation network, and Efi be the CCTV detection image of the drainage pipe in the i-th frame (single-frame input features). The segmentation network achieves accurate localization and feature extraction of defect regions in a single-frame image, providing the basic input for subsequent temporal processing.

[0094] S1-3-2. Based on the defect feature query vector output by the segmentation network, the tracking network is used for dynamic tracking to obtain the defect tracking result output by the tracking network.

[0095] The Tracing Network (Tra) models cross-frame defect tracking as a denoising task, consisting of multiple stacked Transformer denoising blocks. Its core function is to establish a preliminary correspondence between defect instances between adjacent frames. Specifically, in the processing of frame Y, the defect features Cor corrected in frame Y−1 are used... Y−1 As a query, the defect features extracted from the Y-th frame segmentation. As a key value, with noise adder N pairs Defect features generated by adding noise As an ID, a cross-attention mechanism is used to check noisy queries and output stable defect tracking results. The mapping relationship is as follows:

[0096]

[0097] Among them, Q Tra Tra(⋅) is the mapping function of the tracking network to track the defect tracking results output by the tracking network. The tracking network effectively solves the problem of cross-frame correlation of defect instances in long-term video, providing support for continuous defect tracking.

[0098] S1-3-3. Input the defect tracking results output by the tracking network into the correction network for defect calibration, and obtain the defect features optimized by the correction network;

[0099] The Correction (Cor) network takes an aligned sequence of feature representations of the same defect instance as input. It first processes the temporal features using a one-dimensional convolutional network, then introduces a temporal self-attention mechanism. This allows the prediction of the current frame to integrate contextual information from distant frames, thereby blocking the propagation of errors in long temporal detection chains and preventing unexpected interruptions or misalignments in defect identification. The output is an optimized, high-precision defect feature Q. Cor The mapping relationship is as follows:

[0100] Q Cor =Cor(Q Tra Q Seg )

[0101] Here, Cor(⋅) is the mapping function of the correction network. The correction network achieves temporal calibration of defect features and locations, significantly improving the robustness and detection accuracy of the model.

[0102] S1-3-4. Based on the defect features optimized by the modified network, the decoding network is used to decode and obtain the defect repair suggestions output by the decoding network;

[0103] The Decoding (Dec) network decodes the final feature representation of each defect instance into two parallel branches: a detection head and a segmentation head. The detection head is responsible for regressing accurate defect category labels and tight bounding boxes (Bboxes) from the features, while the segmentation head is responsible for generating pixel-level binary masks for the corresponding defects. The mapping relationship is as follows:

[0104] Bbox=Dec(Q Cor )

[0105] Mask=Dec(Q Cor )

[0106] Here, Dec(⋅) is the mapping function of the decoding network. The decoding network completes the final output of defects from features to quantifiable results, providing a data foundation for subsequent defect classification and repair suggestion generation.

[0107] S1-3-5. Based on the defect repair suggestions output by the decoding network, the deep learning algorithm is used to iteratively train the target sample dataset of the drainage pipe to obtain a defect category detection and segmentation model.

[0108] After completing the algorithm architecture, appropriate learning rate, loss function, optimizer and other parameters are set for the model. Using the defect detection and segmentation results output by the decoding network as supervision, a deep learning algorithm that integrates convolutional neural networks and Transformer networks is adopted. Relying on NVIDIA RTX A6000 hardware resources and PyTorch deep learning framework, the processed sample dataset is input into the algorithm for iterative training. The specific parameters are shown in Table 2. During the training process, the loss change is monitored in real time, and the optimal weight is determined based on the loss convergence. Finally, a defect category detection and segmentation model with stable performance and satisfactory accuracy is obtained.

[0109] Table 2

[0110] Parameter Value Parameter Value Dataset proportion 4:1 Batch size 8 Training period (Epoch) 50 Learning rate 0.0001 Weight decay 0.05 Optimizer AdamW

[0111] In summary, steps S1-3-1 to S1-3-5 construct a deep learning model consisting of four cascaded modules: segmentation, tracking, correction, and decoding. The segmentation network extracts single-frame defect features and potential regions through a two-dimensional convolutional network and a Transformer cross-attention mechanism. The tracking network uses denoising modeling to achieve cross-frame defect instance association. The correction network uses one-dimensional convolution and temporal self-attention to calibrate defect features and block error propagation. The decoding network outputs defect categories, bounding boxes, and pixel-level masks in parallel. Leveraging NVIDIA RTX A6000 hardware resources and the PyTorch deep learning framework, the preprocessed target sample dataset is iteratively trained to obtain the final model. This architecture effectively suppresses long-term video noise accumulation and identity drift problems, significantly improving the accuracy and robustness of defect detection, segmentation, and tracking in CCTV videos of drainage pipes, providing stable and reliable technical support for subsequent defect classification and intelligent generation of repair suggestions.

[0112] As one possible implementation, in the above embodiments, step S2 may specifically include the following steps:

[0113] S2-1. The defect detection and segmentation results of the drainage pipe network are structured and organized to obtain the core defect information of the structured organization;

[0114] The results of drainage network defect detection and segmentation are structured and organized, unifying and standardizing multi-dimensional information such as network defect category, defect area ratio, defect level, and corresponding drainage pipe diameter, material, burial depth, service life, road section, construction constraints, and cost constraints. This forms core defect information with complete fields and standardized format, which is used for subsequent matching and combination with instruction templates to ensure a clear, comprehensive, and accurate description of the overall defect status of the drainage network, providing complete and reliable data support for generating standardized instruction texts.

[0115] S2-2. Based on the structured and organized core defect information and the preset instruction template, obtain standardized instruction text;

[0116] The structured defect core information is combined with a preset instruction template to generate semantically coherent text. Semantically coherent text refers to statements that clearly describe the overall defect situation of the drainage pipe network, obtained by combining instructions according to the template. This preset instruction template includes core elements such as task objectives, pipeline basic information, defect detection data, constraints, technical specifications, and output requirements. It can fully incorporate all content, including pipe diameter, material, burial depth, service life, road section, defect type and level, construction restrictions, cost control, and national and industry technical regulations to be followed. After combination according to the template rules, it generates a standardized instruction text that is semantically coherent, logically complete, and clearly defined in its task, which can be directly input into the intelligent agent and accurately understood and executed. Taking a certain pipeline section as an example, the standardized instruction text generated according to the instruction template is as follows:

[0117]

Task Objective

[0118] [Data Input]

[0119] Pipeline basic information: Pipeline diameter [200mm], material [reinforced concrete / cast iron / plastic], burial depth [145.299m], service life [10 years], road section [Heat Source Street]. The pipeline material in this implementation is reinforced concrete.

[0120] Defect detection data: Defect type [crack / deformation / corrosion / misalignment / undulation / disconnection / joint material detachment / branch pipe concealed connection / foreign object penetration / leakage / deposition / scaling / obstacles / residual wall / dam root / tree root / scum], defect level [level 1 / 2 / 3 / 4]. In this embodiment, the defect type is crack and the defect level is 2.

[0121] [Constraints]

[0122] Construction restrictions: Requirements for [nighttime construction / traffic diversion / protection of surrounding buildings] must be met;

[0123] Cost control: The cost of repairing a single defect shall not exceed 20,000 yuan;

[0124] Technical Specifications: Strictly adhere to CJJ 181-2012 "Technical Specification for Inspection and Evaluation of Urban Drainage Pipelines", DB 31 / T444-2009 "Technical Specification for Television and Sonar Inspection and Evaluation of Drainage Pipelines", DB 44 / T 1025-2012 "Technical Specification for Inspection and Evaluation of Urban Public Drainage Pipelines", DB 37 / T 5107-2018 "Technical Specification for Inspection and Evaluation of Urban Drainage Pipelines" and corresponding local drainage network maintenance standards. Furthermore, combine pipeline defect detection reports and historical case studies of similar defect repairs to optimize the proposed solutions, ensuring compliance, feasibility, and adaptability.

[0125]

Output Requirements

[0126] S2-3. Based on the standardized instruction text, the intelligent agent performs autonomous planning and task decomposition to obtain preliminary repair process suggestions.

[0127] Standardized instruction text is input into the agent, which then autonomously plans and breaks down tasks. Through wide-area information retrieval, it outputs structured suggestions for pipeline defect repair. The agent, based on a large language model, possesses functions such as autonomous planning, problem discovery, resource integration, and conclusion generation, as detailed below:

[0128] (1) Autonomous planning: Transform the research problem into a sequence of sub-goals to form an executable plan, and generate a clear roadmap that fits the task requirements. Through proactive planning and systematic reasoning, the limitations of traditional passive retrieval can be effectively overcome.

[0129] This is a specialized technical issue concerning the development of repair solutions for defects in drainage pipe networks. Users need to generate targeted repair suggestions and priority rankings based on given pipe information and defect detection data.

[0130] Analyze user needs:

[0131] 1. Input data: Pipe diameter 200mm, material reinforced concrete, burial depth 145.299m, service life 10 years, heat source street section; defect type crack, level 2;

[0132] 2. Constraints: Nighttime construction, traffic diversion, building protection, cost ≤ 20,000 yuan;

[0133] 3. Technical Specifications: Must comply with multiple technical specifications such as CJJ 181-2012;

[0134] 4. Output requirements: repair plan, priority ranking, and quality acceptance indicators.

[0135] The following information needs to be searched:

[0136] 1. Repair techniques and methods for drainage pipe rupture defects;

[0137] 2. Repair requirements for Level 2 fracture defects as specified in relevant technical specifications such as CJJ 181-2012;

[0138] 3. Methods for estimating the cost of drainage pipe repair;

[0139] 4. Pipeline repair quality acceptance standards.

[0140] (2) Problem identification: Transform the solutions formed in the planning stage into feasible search query combinations, flexibly adjust the search ideas based on the accumulated evidence and the progress of the task, and provide efficient and high-quality guidance for the subsequent resource integration process while taking into account the accuracy of the search and the coverage of information and preventing the loss of key content.

[0141] Search for repair technologies and relevant standards for drainage pipe rupture defects;

[0142] Search for repair techniques for drainage pipe rupture defects, the requirements of the CJJ 181-2012 standard, and repair solutions for level 2 defects;

[0143] Visit the webpage related to trenchless repair technology for drainage pipelines to learn about specific repair techniques and cost control requirements;

[0144] Search for cost estimates for drainage pipe repairs and specific repair plans for Class 2 rupture defects;

[0145] Visit the detailed information page of CJJ 181-2012 standard to learn about the specific repair requirements and quality acceptance standards for Level 2 crack defects;

[0146] Search for specific repair solutions and quality acceptance standards for Class 2 rupture defects in drainage pipes, as well as repair techniques for small-diameter pipes;

[0147] Based on the specific needs of users, it is necessary to focus on searching for trenchless repair technology solutions for small-diameter drainage pipes (200mm), paying particular attention to repair methods, quality acceptance standards, cost control and priority determination standards for level 2 rupture defects;

[0148] Visit the detailed webpage for the UV curing repair quality acceptance standards to obtain the specific requirements for the quality acceptance indicators after repair.

[0149] Search for more information on the criteria for prioritizing the repair of Class 2 rupture defects in 200mm pipes and the selection of repair technologies for small-diameter pipes;

[0150] The search focuses on the technical specifications and cost control analysis of repair technology for Class 2 rupture defects in 200mm small-diameter drainage pipes, with an emphasis on obtaining specific construction parameters, material requirements, equipment configuration, and optimal solution selection under a budget constraint of 20,000 yuan for the point-like in-situ curing method.

[0151] Access detailed technical documents on the trenchless repair construction process of point-based in-situ solidification method to obtain key technical information such as specific construction parameters, material specifications, and equipment configuration;

[0152] Search for more information on cost control, local standard technical requirements, nighttime construction requirements, and priority determination criteria for drainage pipe repair under a budget constraint of 20,000 yuan.

[0153] Access the document on the cost analysis of trenchless repair technology for drainage pipelines to obtain detailed cost structure and control strategies, so as to formulate the optimal solution under the budget constraint of 20,000 yuan;

[0154] Search for more information on nighttime construction requirements, traffic diversion techniques, repair priority criteria, and cost control solutions under a budget constraint of 20,000 yuan.

[0155] (3) Resource integration: It can accurately extract information with relevance and reliability from various resources, and has the ability to navigate autonomously and interact flexibly in complex scenarios. It can adapt to the processing needs of diverse content forms such as text and images. At the same time, it ensures the reliability of the information obtained through strict quality screening and verification mechanisms.

[0156] Based on the information collected, sufficient data is now available to develop a repair plan for a Class 2 rupture defect in a 200mm drainage pipe. Based on the collected technical data, cost analysis, nighttime construction requirements, and priority criteria, a complete repair recommendation, priority ranking, and quality acceptance indicators will be provided to the user.

[0157] (4) Conclusion generation: The scattered search evidence is integrated and sorted out. At the same time, new knowledge is extracted and refined from diverse information. The content is organized according to the established format and logic, and finally a logically coherent and clearly structured conclusion report is generated.

[0158] The HTML content fully responds to users' needs for repair solutions after the detection and segmentation of drainage pipe defects. The content structure is clear and divided into four main parts: defect background and assessment, repair solution design, repair priority ranking, and quality acceptance indicators, which directly correspond to the basic pipe information and defect detection data input by users.

[0159] The defect background section details the pipe diameter of 200mm, the reinforced concrete material, the service life, and the burial depth. Combined with the defect level of 2, it explains the severity and potential hazards of the defect, which meets the requirements of CJJ 181-2012.

[0160] The repair plan addresses the construction constraints (nighttime construction, traffic diversion, and protection of surrounding buildings) and cost control (cost per site not exceeding 20,000 yuan) mentioned by the user. It proposes a trenchless local repair technology—CIPP point-based in-situ solidification method, and provides alternative solutions, construction organization, and economic analysis, demonstrating the targeted nature and feasibility of the plan.

[0161] The remediation priority ranking is scientifically based on defect severity, environmental sensitivity, risk, and resource constraints, ensuring that the priority for meeting user requirements is clear;

[0162] The quality acceptance criteria cover appearance, dimensions, water tightness / air tightness test, hydraulic performance, material performance and safety, and strictly follow relevant technical specifications to ensure controllable repair quality;

[0163] Overall, the content is complete, highly targeted, closely aligned with user needs, meets constraints and technical specifications, and possesses high practical value and professionalism.

[0164] Based on the requirements, a detailed trenchless repair plan has been developed for the Level 2 rupture defect of the DN200 reinforced concrete pipeline on Reyuan Street. Prioritizing structural safety and construction environmental constraints, and ensuring rational resource allocation through scientific prioritization, the plan also provides strict quality acceptance criteria to ensure that the repair results meet industry standards.

[0165] S2-4. Verify the preliminary repair process suggestions and optimize and adjust them in conjunction with the drainage network defect detection and segmentation results to obtain a drainage network defect repair suggestion database.

[0166] The initial repair process recommendations are verified from multiple dimensions, focusing on core aspects such as standard compliance, construction feasibility, cost constraint suitability, and on-site condition matching. For example, it is verified whether the plan complies with multiple national and local technical regulations such as CJJ 181-2012, whether it meets constraints such as nighttime construction, traffic diversion, building protection, and a cost not exceeding 20,000 yuan, and whether it matches the defect type, defect level, and pipeline foundation parameters. Simultaneously, based on the drainage network defect detection and segmentation results, the initial plan is optimized and adjusted accordingly, revising process selection, construction parameters, and priority ranking. Ultimately, a professional, implementable, and compliant drainage network defect repair recommendation is formed and entered into the drainage network defect repair recommendation database, providing reusable standardized solutions for subsequent repairs of similar defects.

[0167] In summary, steps S2-1 to S2-4 combine basic information such as defect category, defect level, pipe diameter, material, burial depth, service life, and road section according to a preset instruction template to generate semantically coherent text that clearly describes the overall defect status of the pipeline network. This text is then input into an intelligent agent based on a large language model. The agent sequentially completes autonomous planning, problem discovery, resource integration, and conclusion generation. Through task decomposition, precise retrieval, standardized matching, and multi-source information verification, it automatically outputs structured repair suggestions containing repair plans, priority ranking, and quality acceptance indicators, while strictly adhering to relevant technical regulations and meeting construction restrictions and cost constraints. The entire process achieves fully automated generation from defect data to professional repair plans, effectively improving the standardization, pertinence, and feasibility of repair suggestions, significantly reducing reliance on manual labor, improving the efficiency of pipeline network defect management, and forming reusable suggestion results, providing efficient and reliable technical support for the intelligent operation and maintenance of urban drainage pipeline networks.

[0168] As one possible implementation, in the above embodiments, step S3 may specifically include the following steps:

[0169] S3-1. Using the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, obtain basic archive information of drainage pipelines;

[0170] Based on the generated database of drainage network defect detection and segmentation results and defect repair suggestions, the corresponding pipeline data is loaded and associated in the spatial visualization interface. At the same time, the pipe diameter, material, burial depth, service life, and road section of the corresponding drainage pipeline are retrieved and integrated to form a complete and standardized basic archive of drainage pipelines, providing a unified and accurate data benchmark for subsequent pipeline data creation, spatial analysis and defect assessment.

[0171] S3-2. Based on the basic archive information of the drainage pipeline, create the current CCTV video data of the drainage pipeline and obtain the drainage pipeline data;

[0172] Based on the basic archive information of drainage pipelines, the CCTV video data of the drainage pipelines currently acquired is matched, associated and data created. First, the current drainage pipeline information is extracted from the video and compared with the archive information to determine whether the pipeline is a new pipeline or an existing pipeline. For new pipelines, the data is drawn on the map and the information is entered. For existing pipelines, the data is supplemented and updated based on the original data. Finally, drainage pipeline data with accurate spatial location and complete attribute information is formed, realizing a one-to-one correspondence between video data and spatial data.

[0173] S3-3. Process and analyze the drainage pipeline data to obtain the associated data from the processing and analysis;

[0174] Systematic processing and spatial analysis are carried out on the completed drainage pipeline data. Data processing operations such as coordinate transformation, data format conversion, and projection settings are performed to ensure data spatial consistency and availability. At the same time, spatial analysis such as overlay analysis and topology analysis are carried out to explore the spatial relationships and potential patterns between pipelines and between pipelines and the surrounding environment. Processed analysis and correlation data including spatial location, topological relationship, and attribute association are obtained to provide spatial support for accurate defect detection and risk assessment.

[0175] S3-4. Based on the associated data of the processing and analysis, the defect category detection and segmentation model is used to obtain the current drainage network defect detection and segmentation results;

[0176] Based on the associated data obtained from the processing and analysis, the pre-trained defect category detection and segmentation model is invoked to automatically detect and segment the current CCTV video data of the drainage pipeline again, outputting the latest defect category, defect area ratio and defect level, so as to obtain real-time and accurate defect detection and segmentation results of the current drainage network, ensuring that the defect status and data are updated synchronously.

[0177] S3-5. Based on the current drainage network defect detection and segmentation results and the drainage pipeline basic file information, perform defect assessment and obtain intelligent generation results of urban drainage network defect repair suggestions.

[0178] The latest obtained results of the current drainage network defect detection and segmentation are combined with the basic archive information of the drainage pipeline to carry out a comprehensive defect assessment. In accordance with relevant technical procedures, the defect hazard assessment, repair priority ranking, repair scheme matching and quality acceptance index determination are completed. Finally, an intelligent generation result of urban drainage network defect repair suggestions is generated, which includes spatial information, defect information, repair suggestions and acceptance standards. It can be intuitively displayed in the GIS platform and recorded in the attribute table to realize the closed loop of the entire process of pipeline defect management.

[0179] In summary, steps S3-1 to S3-5 sequentially integrate defect detection, repair suggestions, and basic pipeline information to form a standardized drainage pipeline basic archive. Then, based on the archive information, CCTV video data is matched and associated with pipeline data creation, automatically identifying and distinguishing between newly added and existing pipelines, and completing the input or update of spatial information. Subsequently, the pipeline data undergoes coordinate transformation, format conversion, projection settings, overlay analysis, and topology analysis to improve data consistency and usability and uncover spatial correlation patterns. On this basis, a defect detection segmentation model is invoked to intelligently identify the video data, obtaining real-time and accurate results of defect categories, proportions, and levels. Finally, a comprehensive assessment is conducted based on the pipeline archive information to complete hazard assessment, priority ranking, scheme matching, and acceptance indicator determination, generating intelligent repair suggestions that can be intuitively displayed on the GIS platform and archived. This process achieves a fully automated closed loop for drainage network data, from file creation, video matching, spatial processing, intelligent detection, and repair decision-making. It effectively improves the standardization of network data management, the real-time nature of defect identification, and the scientific nature of repair recommendations, significantly reduces manual processing costs, and provides efficient and reliable data support and technical assurance for the refined and intelligent operation and maintenance decision-making of urban drainage networks.

[0180] As one possible implementation, in the above embodiments, step S3-2 may specifically include the following steps:

[0181] S3-2-1. Obtain current drainage pipeline information based on the CCTV video data of the current drainage pipeline;

[0182] Based on the basic archive information of drainage pipelines, the CCTV video data of drainage pipelines currently collected is analyzed to extract relevant pipeline information such as pipe diameter, material, spatial location, direction, road section, and video collection characteristics, forming complete information on the current drainage pipelines. This provides a basis for subsequent matching, association, and data creation with archive information.

[0183] S3-2-2. Compare and analyze the current drainage pipeline information with the basic drainage pipeline file information to obtain the comparison result of the current drainage pipeline information;

[0184] The current drainage pipeline information extracted from the video will be matched, associated and compared with the basic archive information of the drainage pipeline item by item. By verifying the spatial characteristics and attribute information of the pipeline, the matching relationship between the two will be clarified, and finally a comparison result will be formed to determine whether the pipeline has been added or not, providing a basis for subsequent classification processing.

[0185] S3-2-3. Based on the comparison results of the current drainage pipeline information, determine whether the current drainage pipeline is a newly added pipeline. If so, proceed to S3-2-4; otherwise, proceed to S3-2-5.

[0186] Based on the comparison results between the current drainage pipeline information and the basic file information, determine whether the target drainage pipeline is a newly added pipeline that is not recorded in the system; if it is determined to be a newly added pipeline, perform the update operation; if it is determined to be an existing pipeline, directly call the corresponding drainage pipeline basic file information.

[0187] S3-2-4. Update the CCTV video set of the drainage pipeline based on the current CCTV video data of the drainage pipeline, and execute S1-1;

[0188] The current drainage pipeline is identified as a newly added pipeline. Based on the current CCTV video data of the drainage pipeline, the CCTV video set of the drainage pipeline is updated and supplemented. At the same time, the spatial drawing and attribute information of the newly added pipeline are completed on the map. Then, the process jumps back to execute step S1-1 to preprocess the video data, and finally forms drainage pipeline data with accurate spatial location and complete attribute information.

[0189] S3-2-5. Based on the current drainage pipeline information, call the basic archive information of the drainage pipeline to obtain the drainage pipeline information;

[0190] The current drainage pipeline is identified as an existing pipeline. Based on the matching information of the current drainage pipeline, the basic file information corresponding to the drainage pipeline is directly retrieved. The pipeline information is supplemented and updated on the basis of the original data to obtain standardized and complete drainage pipeline information, ensuring that the existing pipeline data is accurate and complete, and achieving precise correspondence between video data and spatial data.

[0191] In summary, steps S3-2-1 to S3-2-5 rely on the basic archive information of drainage pipelines. First, pipeline-related information is extracted from the current CCTV video data of the drainage pipelines. Then, it is compared and analyzed item by item with the basic archive information to obtain comparison results for both newly added and existing drainage pipelines. Based on the comparison results, the target drainage pipelines are classified and identified. If it is a newly added pipeline, the CCTV video set is updated, the pipeline is drawn and information is entered on the map, and the video preprocessing step is executed. If it is an existing pipeline, the corresponding basic archive information is directly called and supplemented and updated based on the original data. Finally, drainage pipeline data with accurate spatial location and complete attribute information is formed, achieving a one-to-one correspondence between video data and spatial data. This process can automatically distinguish between newly added and existing pipelines, avoiding duplicate archiving and data redundancy, improving the accuracy and efficiency of pipeline network data matching, and ensuring the completeness and standardization of pipeline spatial and attribute information. This provides reliable data support for subsequent defect detection, assessment, and repair suggestion generation.

[0192] This invention integrates an intelligent method for generating pipeline defect repair suggestions into a pipeline data processing and analysis software system. By loading data into the software, the spatial layout of the drainage pipeline network is displayed intuitively. By opening the attribute table, the attribute information of each pipeline can be obtained. Through functions such as CCTV video path, defect type and level, and defect repair suggestions, the defect category of each pipeline can be detected and segmented, thereby completing the intelligent generation of defect repair suggestions, achieving more precise management, and providing technical support for urban drainage pipeline network defect management.

[0193] Further reference Figure 4 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of an intelligent generation system for urban drainage pipe network defect repair suggestions. This system embodiment is similar to... Figure 1 Corresponding to the method embodiments shown, the system can be specifically applied to various electronic devices.

[0194] like Figure 4 As shown in this embodiment, an intelligent system for generating repair suggestions for urban drainage pipe networks is a GIS software system, which includes: a defect detection and segmentation module, a repair suggestion generation module, and a data processing and analysis module.

[0195] The defect detection and segmentation module is used to collect CCTV video datasets of drainage pipelines and use a defect category detection and segmentation model to obtain defect detection and segmentation results of the drainage pipeline network.

[0196] This module provides the system with intelligent identification of pipeline defects. It is responsible for collecting and accessing CCTV video datasets of drainage pipelines. Through a built-in defect category detection and segmentation model, it automatically identifies, locates and segments pipeline defects in the video at the pixel level, and outputs drainage pipeline defect detection and segmentation results including defect type, defect location, defect level and defect outline range. This provides accurate and standardized defect data support for the generation of subsequent repair suggestions.

[0197] The repair suggestion generation module is used to obtain a database of repair suggestions for drainage network defects based on the drainage network defect detection and segmentation results and a preset instruction template.

[0198] This module is responsible for the intelligent generation of repair solutions and the construction of a knowledge base. Based on the defect detection and segmentation results of the drainage pipe network output by the defect detection and segmentation module, it combines pipeline information, defect information, construction constraints and other contents in a standardized manner with preset instruction templates. Through intelligent agent autonomous planning, information retrieval and solution verification, it generates corresponding repair process suggestions and collects and organizes repair solutions for multiple scenarios and types of defects to build a database of drainage pipe network defect repair suggestions, so as to realize the accumulation and reuse of repair experience.

[0199] The data processing and analysis module is used to obtain intelligent generation results of urban drainage network defect repair suggestions based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database.

[0200] This module is designed for developing urban drainage network data processing and analysis software. It integrates multiple sub-functions, including data loading, data creation, data processing and analysis, data detection and segmentation, and data defect assessment. On one hand, it enables unified management and display of pipeline attributes such as diameter and material through a spatial visualization interface and attribute table for pipeline data. It supports loading, editing, coordinate transformation, projection settings, and spatial analysis such as topology and overlay of pipeline spatial data, ensuring the spatial consistency and usability of pipeline data. On the other hand, it closely links with the pipeline defect category detection and segmentation model and the defect repair suggestion generation algorithm. Users can directly trigger the defect detection and segmentation process of the corresponding pipeline through the attribute table, and automatically call the intelligent generation logic to complete the output of defect repair suggestions based on the detection results. This achieves integrated analysis and business closed loop from pipeline spatial data management and intelligent defect identification to automatic generation of repair solutions.

[0201] In summary, this system mainly consists of a defect detection and segmentation module, a repair suggestion generation module, and a data processing and analysis module. The defect detection and segmentation module collects CCTV video data from drainage pipelines and uses a trained defect category detection and segmentation model to automatically identify, locate, and segment defects at the pixel level, outputting standardized defect detection and segmentation results. The repair suggestion generation module intelligently generates repair process suggestions based on defect detection results and instruction templates, and constructs a reusable defect repair suggestion database. The data processing and analysis module relies on the pipeline data processing and analysis software developed by QGIS, integrating multiple sub-module functions to achieve pipeline attribute management, spatial data processing and analysis, and defect assessment. It supports users in triggering defect detection and repair suggestion generation with a single click through attribute tables. The entire system achieves integrated and intelligent management of drainage pipeline defect detection, spatial data management, and repair suggestion generation, significantly improving defect identification accuracy and solution generation efficiency, simplifying operation and maintenance processes, and providing efficient and reliable technical support for the scientific, refined, and intelligent management of urban drainage pipelines.

[0202] In this embodiment, the specific processing of an intelligent system for generating suggestions for repairing defects in urban drainage pipe networks and the resulting technical effects can be referred to separately. Figure 1 The relevant descriptions of steps S1, S2 and S3 in the corresponding embodiments will not be repeated here.

[0203] In some alternative implementations, the data processing and analysis module is GIS software developed based on the QGIS API interface, such as... Figure 5As shown, it includes a data loading submodule, a data creation submodule, a data processing and analysis submodule, a data detection and segmentation submodule, and a data defect assessment submodule, which comprehensively realize the full lifecycle management of pipeline data and intelligent defect assessment. The specific functions of each submodule are as follows:

[0204] The data loading submodule is used to obtain basic archive information of drainage pipelines by utilizing the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database.

[0205] This submodule, serving as the foundational data access unit for the data processing and analysis module, leverages the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database to integrate and extract key information for each drainage pipeline, including pipe diameter, material, burial depth, service life, road section, defect information, and repair suggestions. This allows for the construction and acquisition of complete and standardized basic archive information for drainage pipelines. Furthermore, this submodule possesses highly compatible data loading and display capabilities, capable of loading and displaying drainage pipeline data in various formats (Shapefile, GeoJSON, Excel, CAD) such as vector data, tabular data, and CAD data. It supports importing multi-source pipeline data into the QGIS-based urban drainage network data processing and analysis software via drag-and-drop or file selection. After loading, the software's spatial visualization interface clearly and intuitively displays the spatial distribution of the drainage network, providing functions such as attribute editing, layer rendering, and field annotation display. Users can perform basic map operations such as zooming in, zooming out, panning, and measurement, providing a unified, accurate, and usable basic data source for the efficient operation of subsequent submodules.

[0206] The data creation submodule is used to create CCTV video data of the current drainage pipeline based on the basic archive information of the drainage pipeline, and to obtain drainage pipeline data;

[0207] This submodule, based on the basic archive information of drainage pipelines obtained by the data loading submodule, performs matching, association, and data creation operations on the currently collected CCTV video data of drainage pipelines. It supports the creation and editing of core pipeline elements such as pipeline points, pipelines, and pipeline zoning surfaces. It can automatically generate and load an attribute table containing eight columns of information, including pipe diameter, material, burial depth, service life, road section, CCTV video path, defect type and level, and defect repair suggestions, in response to the association operation of pipeline layers. Among them, the five columns of data, pipe diameter, material, burial depth, service life, and road section, are automatically assigned and filled, while the three columns of CCTV video path, defect type and level, and defect repair suggestions are initially empty.

[0208] Meanwhile, this submodule can distinguish between newly added pipelines and existing pipelines. For newly added pipelines, it supports digital drawing on the map from scratch and inputting the corresponding attribute information. For existing pipelines, it supports supplementary drawing and attribute improvement and updates based on the original data, ultimately forming drainage pipeline data with accurate spatial location and complete attribute information. This achieves a one-to-one correspondence between CCTV video data and pipeline spatial data, ensuring the integrity, matching, and timeliness of pipeline network data.

[0209] The data processing and analysis submodule is used to process and analyze the drainage pipeline data and obtain related data from the processing and analysis.

[0210] This submodule is responsible for the systematic processing and in-depth analysis of the drainage pipeline data generated by the data creation submodule. It has data processing functions such as coordinate transformation, format conversion, and projection addition, as well as spatial analysis functions such as overlay analysis and topology analysis. It can effectively unify the spatial benchmark and format standard of pipeline data, improve the spatial consistency and usability of drainage pipeline data, and automatically obtain the pipeline basic information and video path information in the attribute table. It can uniformly parse and standardize the associated CCTV video data and pipeline attribute data, and deeply explore the spatial relationships and potential patterns between drainage pipelines and between pipelines and the surrounding environment. This provides high-quality associated data support for subsequent defect detection and assessment, and helps to achieve more accurate pipeline network risk warning and more scientific pipeline network planning and layout.

[0211] The data detection and segmentation submodule is used to obtain the current drainage network defect detection and segmentation results based on the processed and analyzed associated data and the defect category detection and segmentation model.

[0212] This submodule is closely linked to the defect category detection and segmentation model. Based on the processed and analyzed data output by the data processing and analysis submodule, users only need to complete simple video input and parameter settings to trigger the operation of this submodule. It integrates a mature drainage pipe network defect detection and segmentation algorithm. After obtaining the matched CCTV video path information in the attribute table, it automatically calls the trained defect detection and segmentation model to perform refined and fully automatic defect detection and pixel-level segmentation on the current drainage pipe CCTV video data. It intelligently identifies and determines key information such as the pipe network defect category, defect level, defect location, and contour range, and obtains real-time and accurate current drainage pipe network defect detection and segmentation results. The identified defect type and defect level information are automatically and synchronously written into the pipe network attribute table to ensure that the defect data is consistent with the actual state of the pipeline on site, realizing intelligent and automated detection output of pipe network defects.

[0213] The data defect assessment submodule is used to assess defects based on the current drainage network defect detection and segmentation results and the drainage pipeline basic archive information, and to obtain intelligent generation results of urban drainage network defect repair suggestions.

[0214] This submodule, as the core unit of the system's intelligent decision-making, integrates an intelligent agent generation algorithm for defect repair suggestions. Based on the defect detection and segmentation results of the current drainage network output by the data detection and segmentation submodule, and combined with the basic drainage pipeline file information obtained by the data loading submodule, it automatically extracts the pipeline basic information, defect type, and level information already filled in the attribute table. Following a preset instruction template, it completes the standardized combination and encapsulation of information, and then automatically drives the intelligent agent to execute the planning, retrieval, integration, and conclusion generation process. By matching defect types, combining pipeline attributes, and referring to industry standards, it completes the determination of defect severity, priority ranking of repairs, matching of repair schemes, and formulation of quality acceptance indicators. Finally, it intelligently outputs structured and implementable intelligent generation results of urban drainage network defect repair suggestions, and automatically writes the final generated repair suggestion information into the pipeline attribute table, realizing the fully automated output of defect repair suggestions and providing accurate and efficient decision support for pipeline network operation and maintenance.

[0215] In summary, the data processing and analysis module of this system comprises five sub-modules: data loading, data creation, data processing and analysis, data detection and segmentation, and data defect assessment. The data loading sub-module is compatible with various formats of pipeline data loading and display, and provides functions such as attribute editing and map operations. It combines defect detection and repair suggestion data to form a complete basic archive of drainage pipeline information. The data creation sub-module can match and create data based on archive information using CCTV video data, supporting the drawing of new pipelines and the supplementation and updating of existing pipelines. The data processing and analysis sub-module improves the spatial consistency and usability of data through functions such as coordinate transformation, projection settings, overlay, and topology analysis, uncovering spatial patterns to support risk warning and planning layout. The data detection and segmentation sub-module integrates defect detection and segmentation algorithms, enabling intelligent output of pipeline network defect categories and levels through simple operation. The data defect assessment sub-module relies on intelligent agent algorithms to combine detection results and archive information to complete defect assessment, achieving intelligent output of repair suggestions. The system achieves a closed-loop process encompassing unified management of multi-source pipeline network data, standardized processing of spatial data, and intelligent generation of automatic defect detection and repair solutions. This significantly improves the standardization of urban drainage pipeline network data management, the accuracy of defect identification, and the efficiency of operation and maintenance decisions, providing efficient and reliable support for the refined, scientific, and intelligent governance of pipeline networks.

[0216] It should be noted that the implementation details and technical effects of each module and unit in the device provided in the embodiments of this disclosure can be referred to the description of other embodiments in this disclosure, and will not be repeated here.

[0217] The following is for reference. Figure 6It shows a schematic diagram of the structure of a computer system 500 suitable for implementing the electronic device of the present disclosure. Figure 6 The computer system 500 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0218] like Figure 6 As shown, the computer system 500 may include a processing device 501 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in ROM 502 or a program loaded from storage device 508 into random access RAM 503. RAM 503 also stores various programs and data required for the operation of the computer system 500. The processing device 501, ROM 502, and RAM 503 are interconnected via bus 504. I / O interface 505 is also connected to bus 504.

[0219] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows computer system 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 A computer system 500 with various electronic devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0220] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, it performs the functions defined in the methods of embodiments of this disclosure.

[0221] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0222] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0223] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the following functions: Figure 1 The illustrated embodiments and their alternative implementations demonstrate a method for intelligently generating suggestions for repairing defects in urban drainage pipe networks.

[0224] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0225] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0226] The units or modules involved in the embodiments described in this disclosure can be implemented in software or hardware. The names of the units or modules do not necessarily limit the unit itself; for example, an acquisition module can also be described as "acquiring preset prompts, including modality fusion prompts, attention mechanism prompts, and / or time-related prompts."

[0227] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for intelligently generating suggestions for repairing defects in urban drainage pipe networks, characterized in that, include: S1. Collect CCTV video dataset of drainage pipelines and use the defect category detection and segmentation model to obtain the defect detection and segmentation results of drainage pipeline network; S2. Based on the drainage network defect detection and segmentation results and the preset instruction template, obtain a drainage network defect repair suggestion database; S3. Based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, obtain the intelligent generation result of urban drainage network defect repair suggestions.

2. The intelligent generation method for urban drainage pipe network defect repair suggestions according to claim 1, characterized in that, S1. Collect CCTV video dataset of drainage pipelines and use a defect category detection and segmentation model to obtain defect detection and segmentation results for the drainage pipeline network, including: Data preprocessing of CCTV video sets of drainage pipes is performed using the defect categories of drainage pipe networks to obtain target video segments of drainage pipes; The target video segments of the drainage pipeline are labeled according to the frame-by-frame annotation density to establish a target sample dataset of the drainage pipeline; Based on the target sample dataset of the drainage pipe, a deep learning algorithm is used to train and construct a defect category detection and segmentation model. The deep learning algorithm includes a segmentation network, a tracking network, a correction network, and a decoding network. The defect category detection and segmentation model is used to detect and segment the CCTV video set of the drainage pipe to obtain the defect category and the area ratio of the drainage pipe defect. Based on the proportion of the defect area in the drainage pipe, the defect level of the drainage pipe is determined, and by combining the defect category of the drainage pipe with the proportion of the defect area in the drainage pipe, the defect detection and segmentation results of the drainage pipe network are obtained.

3. The intelligent generation method for urban drainage pipe network defect repair suggestions according to claim 2, characterized in that, Based on the target sample dataset of the drainage pipes, a deep learning algorithm is used to train a defect category detection and segmentation model, including: Input the target sample dataset of the drainage pipe into the segmentation network to obtain the defect feature query vector output by the segmentation network; Based on the defect feature query vector output by the segmentation network, the tracking network is used for dynamic tracking to obtain the defect tracking result output by the tracking network; The defect tracking results output by the tracking network are input into the correction network for defect calibration, and the defect features optimized by the correction network are obtained. Based on the defect features optimized by the modified network, the decoding network is used to decode and obtain defect repair suggestions output by the decoding network. Based on the defect repair suggestions output by the decoding network, the deep learning algorithm is used to iteratively train the target sample dataset of the drainage pipe to obtain a defect category detection and segmentation model.

4. The intelligent generation method for urban drainage pipe network defect repair suggestions according to claim 1, characterized in that, S2. Based on the drainage network defect detection and segmentation results combined with a preset instruction template, obtain a drainage network defect repair suggestion database, including: The defect detection and segmentation results of the drainage pipe network are structured and organized to obtain the core defect information. Based on the structured and organized core defect information combined with the preset instruction template, standardized instruction text is obtained; Based on the standardized instruction text, the intelligent agent performs autonomous planning and task decomposition to obtain preliminary repair process suggestions. The initial repair process suggestions are verified and optimized based on the drainage network defect detection and segmentation results to obtain a drainage network defect repair suggestion database.

5. The intelligent generation method for urban drainage pipe network defect repair suggestions according to claim 1, characterized in that, S3. Based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, obtain the intelligent generation results of urban drainage network defect repair suggestions, including: Using the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database, basic archive information of drainage pipelines is obtained; Based on the basic archive information of the drainage pipeline, CCTV video data of the current drainage pipeline is created to obtain drainage pipeline data; The drainage pipeline data is processed and analyzed to obtain related data from the processing and analysis; Based on the associated data processed and analyzed, the defect category detection and segmentation model is used to obtain the current drainage network defect detection and segmentation results. Based on the current drainage network defect detection and segmentation results and the drainage pipeline basic file information, a defect assessment is performed to obtain intelligent generation results of urban drainage network defect repair suggestions.

6. The intelligent generation method for urban drainage pipe network defect repair suggestions according to claim 5, characterized in that, Based on the basic archive information of the drainage pipeline, CCTV video data of the current drainage pipeline is created, and drainage pipeline data is obtained, including: Based on the CCTV video data of the current drainage pipeline, obtain the current drainage pipeline information; The current drainage pipeline information is compared and analyzed with the basic drainage pipeline file information to obtain the comparison result of the current drainage pipeline information; Based on the comparison results of the current drainage pipeline information, determine whether the current drainage pipeline is a newly added pipeline. If it is, execute the first operation; otherwise, execute the second operation. The first operation is: updating the CCTV video set of the drainage pipeline based on the current CCTV video data of the drainage pipeline, and performing the third operation; The second operation is: to retrieve the drainage pipeline basic file information based on the current drainage pipeline information to obtain the drainage pipeline information; The third operation is to preprocess the CCTV video set of drainage pipes using the drainage pipe network defect category to obtain the target video segment of the drainage pipes.

7. An intelligent system for generating suggestions for repairing defects in urban drainage pipe networks, employing the method described in any one of claims 1-6, characterized in that, include: Defect detection and segmentation module, repair suggestion generation module, and data processing and analysis module; The defect detection and segmentation module is used to collect CCTV video datasets of drainage pipelines and use a defect category detection and segmentation model to obtain defect detection and segmentation results of the drainage pipeline network. The repair suggestion generation module is used to obtain a database of repair suggestions for drainage network defects based on the drainage network defect detection and segmentation results and a preset instruction template. The data processing and analysis module is used to obtain intelligent generation results of urban drainage network defect repair suggestions based on the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database.

8. The intelligent generation system for urban drainage pipe network defect repair suggestions according to claim 7, characterized in that, The data processing and analysis module includes a data loading submodule, a data creation submodule, a data processing and analysis submodule, a data detection and segmentation submodule, and a data defect assessment submodule. The data loading submodule is used to obtain basic archive information of drainage pipelines by utilizing the drainage network defect detection and segmentation results and the drainage network defect repair suggestion database. The data creation submodule is used to create CCTV video data of the current drainage pipeline based on the basic archive information of the drainage pipeline, and to obtain drainage pipeline data; The data processing and analysis submodule is used to process and analyze the drainage pipeline data and obtain related data from the processing and analysis. The data detection and segmentation submodule is used to obtain the current drainage network defect detection and segmentation results based on the processed and analyzed associated data and the defect category detection and segmentation model. The data defect assessment submodule is used to assess defects based on the current drainage network defect detection and segmentation results combined with the drainage pipeline basic archive information, and to obtain intelligent generation results of urban drainage network defect repair suggestions.

9. An electronic device, characterized in that, include: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any one of claims 1-7.

Images