Tunnel drainage system dredging engineering data intelligent acquisition and processing method and system

Abstract

This invention relates to the field of engineering data processing, and more particularly to an intelligent data acquisition and processing method and system for dredging engineering projects in tunnel drainage systems. The method includes the following steps: receiving processing tasks based on a cloud-based dredging management platform; setting acquisition constraints based on the processing tasks to obtain an acquisition constraint set; driving a pipeline detection robot to perform operational detection based on the acquisition constraint set, outputting a first dataset; adaptively optimizing the first dataset to output a second dataset; performing index-based analysis on the second dataset to output multi-dimensional analysis indicators; and planning the dredging operation path based on the multi-dimensional analysis indicators, synchronously uploading the data to the cloud-based dredging management platform to complete the data acquisition and analysis operation. This invention optimizes dredging operation planning, improves dredging efficiency, and enhances the accurate structural analysis of the drainage system in the early stages through precise data acquisition and processing.

Description

Technical Field

[0001] This invention relates to the field of engineering data processing, and in particular to a method and system for intelligent acquisition and processing of dredging engineering data for tunnel drainage systems. Background Technology

[0002] With the development of the Internet of Things, sensor technology, and intelligent robotics, pipeline detection equipment and data acquisition methods have been rapidly applied. However, traditional inspection equipment often only collects video or water flow parameters, resulting in isolated data that lacks real-time processing capabilities, making it difficult to form a complete pipeline operation status assessment system. Furthermore, the complex internal environment of pipelines, including water flow fluctuations, sediment buildup, and structural anomalies, makes it impossible to accurately determine pipeline patency and structural safety issues from a single data source, increasing the difficulty of decision-making for dredging operations.

[0003] While some intelligent inspection methods can collect data from multiple sources, they largely rely on manual intervention for data processing and analysis, lacking automated and real-time processing capabilities, making it difficult to meet the rapid maintenance needs of large-scale pipeline networks. Furthermore, existing data processing methods lack quantitative analysis tools for indicators such as the degree of siltation, cross-sectional blockage rate, and structural damage, hindering the scientific and precise planning of pipeline network risk assessments and dredging operation routes. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes an intelligent data acquisition and processing method and system for dredging engineering in tunnel drainage systems, thereby resolving at least one of the aforementioned technical issues.

[0005] To achieve the above objectives, this invention provides a method for intelligent data acquisition and processing of dredging projects in tunnel drainage systems, comprising the following steps: Step S1: Receive and process tasks based on the cloud-based dredging management platform; set collection constraints based on the processing tasks to obtain a collection constraint set; Step S2: Drive the pipeline detection robot to perform operational detection based on the collected constraint set, and output the first dataset; Step S3: Perform adaptive optimization on the first dataset and output the second dataset; Step S4: Perform index-based analysis on the second dataset and output multidimensional analysis indicators; Step S5: Plan the dredging operation path based on multi-dimensional analysis indicators, and upload it to the cloud-based dredging management platform to complete the data collection and analysis.

[0006] This specification provides an intelligent data acquisition and processing system for tunnel drainage system dredging projects, used to execute the intelligent data acquisition and processing method for tunnel drainage system dredging projects as described above, including: The data acquisition and setting module is used to receive and process tasks based on the cloud-based dredging and management platform; and to set data acquisition constraints based on the processing tasks to obtain a set of data acquisition constraints. The task detection module is used to drive the pipeline detection robot to perform task detection based on the collected constraint set and output the first dataset. The data optimization module is used to adaptively optimize the first dataset and output the second dataset. The analysis module is used to perform index-based analysis on the second dataset and output multidimensional analysis indicators. The planning module is used to plan dredging operation paths based on multi-dimensional analysis indicators and upload them synchronously to the cloud-based dredging management platform to complete the data collection and analysis.

[0007] The specific benefits of this invention are as follows: By receiving tasks and setting collection constraints through a cloud platform, standardized operations, unified parameters, and efficient data distribution can be achieved, improving the consistency and accuracy of data collection and providing a reliable foundation for subsequent analysis. Using collection constraints to drive robot inspections enables continuous acquisition of high-quality video, sensor, and positioning data, ensuring data integrity and spatial continuity while reducing the risks of manual operations and improving operational efficiency. Through image optimization, signal smoothing, and positioning compensation, adaptive optimization improves data quality, eliminates noise and anomalies, and makes subsequent indicator calculations more accurate and reliable. Indicator-based analysis transforms multi-source data into quantifiable indicators, such as siltation depth, blockage rate, and patency coefficient, enabling visualization and scientific evaluation of pipeline operation status and providing a basis for operational decisions. Route planning based on indicators and uploading to the cloud allows for prioritizing high-risk areas, improving dredging efficiency, while simultaneously achieving data sharing and closed-loop operational management, enhancing the scientific nature of decision-making. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the steps of an intelligent data acquisition and processing method for dredging engineering in a tunnel drainage system according to the present invention. Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1. Figure 3 This is a flowchart illustrating the detailed implementation steps of step S2. Detailed Implementation

[0009] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0010] This application provides a method and system for intelligent data acquisition and processing of dredging engineering data in a tunnel drainage system. The executing entities of the method and system include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, and network upload devices that can be considered general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0011] Please see Figures 1 to 3 This invention provides a method for intelligent data acquisition and processing of dredging projects in tunnel drainage systems, comprising the following steps: Step S1: Receive and process tasks based on the cloud-based dredging management platform; set collection constraints based on the processing tasks to obtain a collection constraint set; Step S2: Drive the pipeline detection robot to perform operational detection based on the collected constraint set, and output the first dataset; Step S3: Perform adaptive optimization on the first dataset and output the second dataset; Step S4: Perform index-based analysis on the second dataset and output multidimensional analysis indicators; Step S5: Plan the dredging operation path based on multi-dimensional analysis indicators, and upload it to the cloud-based dredging management platform to complete the data collection and analysis.

[0012] In the embodiments of the present invention, see Figure 1 This is a flowchart illustrating the steps of an intelligent data acquisition and processing method for dredging projects in a tunnel drainage system according to the present invention. In this example, the steps of the intelligent data acquisition and processing method for dredging projects in a tunnel drainage system include: Step S1: Receive and process tasks based on the cloud-based dredging management platform; set collection constraints based on the processing tasks to obtain a collection constraint set; In this embodiment, maintenance personnel create work tasks and input relevant information on the cloud platform, such as tunnel name, work section, drainage pipeline number, estimated work time, and type of work equipment. After the task is created, the platform generates a unique task number and writes the task information into the database. The on-site work terminal establishes a connection with the cloud platform through a network communication interface and queries the platform for work task information at a preset cycle, such as sending a task request every 30 seconds. When the terminal receives the corresponding task, it parses the task data and extracts information such as pipeline structure information, work section length, and historical maintenance records involved in the task. Based on this information, data collection constraints are set, including parameters such as video acquisition resolution, sensor sampling frequency, positioning accuracy, and data sampling interval. For example, when the pipe diameter is less than DN300, the video acquisition resolution is set to 1920×1080 and the video frame rate is set to 25fps; when the pipe diameter is greater than DN500, the resolution can be appropriately reduced to 1280×720 to reduce data storage pressure. At the same time, the sensor sampling frequency is set according to the water flow monitoring requirements, such as the sampling frequency of flow, water level, and water pressure sensors being set to the range of 5Hz to 10Hz. For the positioning module, the position update interval is set to 0.2s to 0.5s to keep the positioning accuracy within ±0.2m. Additionally, the video keyframe extraction interval needs to be set, for example, extracting one key image frame every 1 second for subsequent image recognition processing.

[0013] Step S2: Drive the pipeline detection robot to perform operational detection based on the collected constraint set, and output the first dataset; In this embodiment, the constraint parameters are loaded into the robot control terminal, and the industrial camera, multimodal sensors, and positioning module inside the robot are uniformly configured. The robot is placed inside the drainage pipe through the manhole inlet, and begins to move and inspect according to the preset inspection path. The robot's movement speed is typically controlled between 0.1 m / s and 0.3 m / s to ensure stable video footage and clear recording of the pipe's interior. During the inspection, the industrial camera continuously acquires video data from inside the pipe and records the video footage at a frame rate of 25 fps, while extracting a key image frame every 1 second. The multimodal sensors synchronously acquire water flow-related parameters, such as flow rate, water level, and water pressure. The flow rate sensor typically measures from 0 m³ / h to 50 m³ / h, the water level from 0 m to 1.5 m, and the water pressure from 0 MPa to 0.5 MPa. Each sensor records data at a sampling frequency of 5 Hz to 10 Hz. Simultaneously, the positioning module records the robot's spatial coordinates within the pipe at 0.2-second intervals, thus forming a continuous motion trajectory. For example, when inspecting a pipe approximately 40 meters long, the entire inspection process lasts about 200 seconds, during which approximately 5,000 frames of video images, approximately 1,000 water flow parameter data points, and approximately 1,000 spatial location records can be collected. All collected data is stored in chronological order in the robot's mobile terminal, forming a complete raw dataset, which is the first dataset.

[0014] Step S3: Perform adaptive optimization on the first dataset and output the second dataset; In this embodiment, image preprocessing is performed on the video image data, including image denoising and contrast enhancement. Image denoising can employ spatial filtering methods, such as median filtering using a 3×3 or 5×5 neighborhood window, to eliminate random noise in the image. Gray-level histogram equalization is used to enhance image brightness, making the details of the pipe's inner wall structure clearer. Next, the water flow parameter data undergoes signal smoothing, using a moving average method to reduce data fluctuations. For example, a moving window length of 5-10 sampling points is set, and the data within the window is averaged to obtain smoothed flow rate, water level, and water pressure change curves. Simultaneously, outlier identification is performed on the smoothed data; for example, when a data point deviates from the average by more than three times the standard deviation, it is marked as an outlier and replaced using linear interpolation to maintain the continuity of the data sequence. Furthermore, drift analysis is required for the positioning data; by comparing the robot's recorded position with the known structural position of the pipe, positioning error parameters are calculated. For example, when inspecting a 40m pipeline, the positioning error may gradually accumulate to 0.25m. By correcting the position data using error compensation methods, the final positioning error can be controlled within ±0.1m. After image optimization, signal smoothing, and positioning compensation processing, a more stable and accurate dataset can be obtained, which is the second dataset.

[0015] Step S4: Perform index-based analysis on the second dataset and output multidimensional analysis indicators; In this embodiment, visual recognition analysis is performed on video image data. Image segmentation and texture feature analysis methods are used to identify deposition areas inside the pipe and calculate the deposition height, thus obtaining a sedimentation depth index. Simultaneously, the cross-sectional blockage rate is calculated based on the area ratio of the deposition area within the pipe cross-section. For example, in a DN400 pipe, when the deposition height is approximately 80mm, the cross-sectional blockage rate may reach around 0.30. Next, pipe structural features in the video images are detected, such as identifying cracks, misaligned joints, and pipe wall deformation. The degree of structural damage is assessed based on parameters such as crack length and width. For example, when the crack length exceeds 100mm or the width is greater than 3mm, it can be determined as relatively severe structural damage. Furthermore, operational status analysis is performed in conjunction with water flow parameters. The flow rate coefficient is calculated by analyzing the relationship between flow rate and water level. For example, when the actual flow rate is 14m³ / h and the design flow rate is 20m³ / h, the flow utilization rate is approximately 0.70, thus the flow rate coefficient is approximately 0.70. By comprehensively analyzing these indicators, a multidimensional set of analytical indicators can be formed, including key indicators such as siltation depth, cross-sectional blockage rate, structural damage level, and water flow coefficient, which are used to reflect the overall operating status of drainage pipelines.

[0016] Step S5: Plan the dredging operation path based on multi-dimensional analysis indicators, and upload it to the cloud-based dredging management platform to complete the data collection and analysis.

[0017] In this embodiment, risk assessments are conducted at various locations along the pipeline based on indicators such as siltation depth, cross-sectional blockage rate, and water flow coefficient, and the risk level is divided into multiple grades. For example, a cross-sectional blockage rate below 20% is considered a low-risk area, 20% to 50% is a medium-risk area, and above 50% is a high-risk area. These risk grades are mapped onto the pipeline topology to form a complete blockage risk distribution map. The dredging operation path is planned according to the risk distribution, for example, prioritizing high-risk areas and connecting adjacent high-risk pipe sections into a continuous operation path to reduce equipment travel distance and operation time. The path planning process also needs to consider equipment movement speed and operation efficiency; for example, when the dredging equipment moves at a speed of 0.2 m / s, cleaning a 10 m section of severely silted area may take approximately 10 minutes. By comprehensively analyzing the risk level and spatial location, an optimal dredging operation path can be generated, providing the operation sequence and estimated completion time. Finally, the detection data, multi-dimensional analysis indicators, blockage risk distribution map, and dredging operation path information are compiled into a comprehensive analysis report, which is then synchronously uploaded to the cloud-based dredging management platform via a network communication interface. Upon receiving the report, the platform completes the entire process of data collection, analysis, and operation planning, thereby achieving intelligent management of tunnel drainage system dredging projects.

[0018] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Tasks are received and processed based on a cloud-based dredging management platform; Based on the processing task, relevant design documents are extracted; the relevant design documents are analyzed for pipeline topology to extract pipe segment connection relationships, node locations and pipe diameter parameters, and generate pipeline topology information; the relevant design documents include tunnel names, sections, drainage pipeline numbers, manhole locations, pipe diameter parameters, historical maintenance records and design drawings.

[0019] Based on the relevant design documents, information analysis was performed to obtain basic pipeline information; Based on the pipeline basic information and pipeline topology information, data acquisition constraints are set to obtain a data acquisition constraint set; the data acquisition constraint set includes video resolution, sensor sampling frequency, positioning module accuracy, and data sampling interval.

[0020] In this embodiment, a cloud-based dredging management platform is used to uniformly receive and manage on-site work tasks. The cloud platform is typically deployed on an operations and maintenance center server or in a private cloud environment, interacting with on-site mobile terminals, detection equipment, and work control terminals via network communication interfaces. When operations and maintenance personnel create a dredging and detection task on the platform, a unique task number is generated, and relevant task information is recorded, including the tunnel name, work section number, drainage pipeline number, planned work time, and equipment type. On-site work terminals send task query requests to the cloud platform at preset time intervals, typically 30-60 seconds. When the platform detects a corresponding task, it encapsulates the task data in a structured data format and sends it to the work terminal. After receiving the task, the terminal performs task validity verification and equipment matching verification, such as verifying whether the video module of the acquisition equipment meets the 1920×1080 resolution requirement, whether the positioning module accuracy is better than ±0.5m, and whether the number of sensor acquisition channels is no less than four. Once the equipment parameters meet the operational requirements, the work terminal establishes a task execution session and completes the initialization of the acquisition equipment, including the startup and parameter loading of the video acquisition unit, positioning module, and environmental monitoring sensors. Simultaneously, it records information such as the task start time, equipment number, and operator number, and uploads this information to a cloud database for storage. When network bandwidth conditions are unstable, video encoding compression methods can be used to reduce the transmission load; for example, using H.265 encoding to control the video bitrate within the range of 4Mbps to 8Mbps ensures the stability of data transmission.

[0021] Based on the task information, a file index is established. Design data related to the drainage system in this section is extracted through database retrieval, and pipeline information, manhole location data, and pipe diameter parameters are read from the drawings. The drainage pipeline layer in the CAD drawings is parsed to extract pipeline centerline data and line segment endpoint coordinates. To establish the pipeline network structure, manholes or pipe connection points in the drainage system are defined as nodes, and drainage pipe segments between two nodes are defined as connecting edges, thus constructing the drainage network topology. During the parsing process, line segment endpoints are identified using a spatial coordinate matching method. When the spatial distance between two endpoints is less than 0.3m, they are considered the same connecting node and merged to form a complete node set. Each node records its spatial coordinate information, while each pipe segment records attributes such as the starting node number, ending node number, pipe diameter parameters, and pipe segment length. For example, the diameter of drainage pipes is typically in the range of DN200 to DN800, and the length of a single pipe segment is generally between 20m and 50m. To ensure the accuracy of the topology, a consistency check is also performed on the parsing results, such as detecting the existence of isolated nodes, duplicate pipe segments, or abnormal connection relationships. When abnormalities are found, such as pipe section length less than 0.5m or pipe diameter less than DN100, the relevant data should be marked for subsequent verification.

[0022] The pipeline attribute information in the design drawings is analyzed, including parameters such as pipeline material, design slope, design flow rate, and construction time. Simultaneously, maintenance information corresponding to the pipe segment is retrieved from historical maintenance records, such as historical dredging time, blockage type, number of repairs, and maintenance cycle. This data is then uniformly linked according to the pipe segment number, integrating information from different data sources into a single pipe segment record to form complete pipeline attribute data. For example, a drainage pipe segment can be recorded as: Pipeline number P-1056, pipe diameter DN400, pipe material reinforced concrete, design slope 0.7%, pipe segment length approximately 38m, construction year 2016, and 4 historical maintenance visits. Further statistical analysis is performed on this pipeline attribute data, such as calculating the distribution ratio of pipe segments with different diameters in the entire drainage network and analyzing the relationship between slope conditions and blockage risk. Generally, drainage pipe segments with diameters in the DN300-DN500 range account for a higher proportion, while segments with smaller slopes are more prone to siltation problems. In addition, spatial analysis can be performed on the location of manholes. The distance between adjacent manholes can be calculated using GIS coordinates. Generally, the distance between manholes is about 30m to 50m. When the distance between manholes exceeds 60m, the drainage pipes in this section are more prone to sedimentation or blockage. Therefore, this section can be marked as a key monitoring area.

[0023] The data acquisition constraints mainly involve video acquisition parameters, sensor sampling frequency, positioning accuracy, and data sampling time intervals. The video acquisition resolution is determined based on the pipe diameter. When the pipe diameter is less than DN300, the video acquisition resolution is set to 1920×1080, and the video frame rate is set to 25fps to clearly identify pipe wall cracks, sediment, or small obstacles. When the pipe diameter is greater than DN500, the resolution can be adjusted to 1280×720 to reduce video data storage and transmission pressure. For environmental monitoring sensors, such as water level sensors, flow velocity sensors, and gas concentration detection devices, the sampling frequency is set according to the complexity of the pipeline environment. In ordinary pipe sections, the sampling frequency can be set to 5Hz, while in key pipe sections prone to siltation or blockage, it can be increased to 10Hz to improve the accuracy of data change monitoring. Regarding the positioning module, an inertial navigation and odometer fusion positioning method can be used, with positioning accuracy generally controlled within ±0.2m. When the pipe section length exceeds 40m, the location of the inspection well can be used as a positioning correction point to correct accumulated errors. In addition, the data sampling time interval needs to be set. For example, the sensor data sampling interval can be set to 0.2s to 0.5s. At key node locations (such as manholes or locations where pipe diameter changes), the sampling interval can be shortened to 0.1s to obtain higher accuracy data records.

[0024] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Standardized operation configurations are set based on the collection constraint set; The pipeline detection robot is driven to perform detection operations and extract the detection dataset based on standardized operating configurations. Perform data format conversion and timeline synchronization alignment on the probe dataset to output the first dataset.

[0025] In this embodiment, based on parameters such as video resolution, video frame rate, sensor sampling frequency, positioning accuracy, and data sampling interval set in the acquisition constraint set, the parameters of each acquisition module of the pipeline detection equipment are uniformly configured. The video acquisition module is configured by setting the video acquisition resolution according to the pipe diameter and detection requirements. For example, when the drainage pipe diameter is less than DN300, the video resolution is set to 1920×1080 and the frame rate to 25fps; when the pipe diameter is greater than DN500, the resolution can be adjusted to 1280×720 to reduce data storage and transmission pressure. The sensor acquisition modules, including water level monitoring sensors, flow velocity sensors, and gas concentration detection devices, are configured by setting the sensor sampling frequency to 5Hz~10Hz according to the settings in the acquisition constraint set, and a unified data output format is adopted, such as using a floating-point format and retaining two or three decimal places of precision. For the positioning module, the positioning error is controlled within ±0.2m according to the operational accuracy requirements, and several positioning correction points are set in the detection path, for example, a positioning correction is set every 30m~40m to reduce cumulative error. Regarding data recording, a unified data sampling time interval needs to be set. For example, the video keyframe recording interval should be set to 1 second, and the sensor data sampling interval should be set to 0.2 to 0.5 seconds. In areas with varying manhole diameters or pipe diameters, the sampling interval should be shortened to 0.1 seconds to improve data accuracy in critical areas. Furthermore, a unified definition of the data storage structure is required. For instance, video data should be stored in MP4 format, sensor data should be recorded in CSV or JSON format, and positioning data should be recorded using timestamps and spatial coordinates.

[0026] After completing the standardized operation configuration, the pipeline inspection robot is driven by this configuration to enter the drainage pipeline and perform inspection operations. At the start of the operation, the robot enters the drainage pipeline through the manhole inlet and moves along the path determined by the pipeline topology information. The robot is typically equipped with a high-definition camera, environmental monitoring sensors, and a positioning module, collecting information about the internal environment of the pipeline through multi-sensor collaboration. During movement, the robot performs inspections at a set speed, typically controlled within the range of 0.1 m / s to 0.3 m / s, to ensure stable video footage and clear recording of the pipe wall conditions. The video acquisition module continuously records images inside the pipeline and generates a video data stream at a frame rate of 25 fps, while extracting a key image every second for subsequent rapid retrieval and defect identification. Simultaneously, environmental monitoring sensors collect data at preset frequencies; for example, water level and flow velocity sensors record drainage status at 5 Hz, while gas concentration detection devices monitor the gas environment inside the pipeline at 10 Hz. The positioning module records the robot's trajectory within the pipeline using a combination of odometry and inertial measurement units, recording its position coordinates every 0.2 seconds to form continuous spatial position information. When the robot passes manholes or areas of pipe diameter change, the positioning information is corrected to reduce accumulated errors. All collected data, including video data, keyframe images, sensor monitoring data, and positioning trajectory data, is initially cached within the robot and recorded in a uniform chronological order. For example, during the inspection of a pipeline approximately 40 meters long, the video data volume is approximately 1GB, while approximately 200 sensor data records and approximately 200 positioning trajectory data records are generated.

[0027] The data formats of video, image, sensor, and positioning data from the detection data undergo conversion. Video data is typically stored in H.264 or H.265 encoding, while keyframe image data may exist in JPEG or PNG format. To unify the data structure, keyframe images are uniformly converted to JPEG format and named according to a unified naming convention, such as "timestamp + location number". Sensor data is uniformly converted to structured tabular data, such as CSV format, and its content is recorded according to a unified field structure, typically including timestamp, sensor number, sample value, and unit information. Positioning data is uniformly converted to time-series coordinate data, such as recording it using a "timestamp—X coordinate—Y coordinate—Z coordinate" data structure. After format unification, timeline synchronization processing is required for all types of data. Since the time intervals for video acquisition, sensor acquisition, and positioning acquisition are different, a unified processing method using timestamp alignment is necessary. The specific approach involves using the keyframe timestamps of the video as the baseline timeline and performing interpolation or matching processing on the sensor data and positioning data. For example, when the sensor data sampling interval is 0.2 seconds and the keyframe interval is 1 second, the sensor data record closest to the keyframe time can be selected through time matching. When there is a deviation between the positioning data and the keyframe time, the position coordinates at the corresponding time point can be calculated through linear interpolation, thereby achieving time series unification. After data format conversion and timeline synchronization processing, a data set with a unified structure can be formed, where each record contains information such as the video keyframe at the corresponding time point, sensor monitoring data, and the robot's spatial position.

[0028] In this embodiment, the job detection specifically includes; The pipeline inspection robot device includes an industrial camera, a multimodal sensor, and a mobile terminal; The configuration of the pipeline inspection robot is adjusted according to the standardized operation configuration described above; Using industrial cameras to capture video of the inside of pipes; Water flow parameters are obtained by collecting flow rate, water level, and water pressure in the pipeline using multimodal sensors. Real-time spatial location is collected based on the positioning module; The system uses mobile terminals to locally store videos of the pipeline interior, water flow parameters, and real-time spatial location, thus constructing a detection dataset.

[0029] In this embodiment, the pipeline inspection robot typically consists of a mobile chassis, an industrial camera, a multimodal sensor assembly, a positioning module, and a mobile terminal. Each module is centrally scheduled via an internal control unit. During configuration adjustments, a standardized operation configuration file is imported into the mobile terminal, and configuration commands are sent to the robot control unit via the control interface. Parameters are loaded for the video acquisition module, such as setting the industrial camera's video resolution to 1920×1080 or 1280×720 and the video frame rate to 25fps. Simultaneously, the exposure time and gain parameters are adjusted according to the lighting conditions inside the pipeline; for example, the exposure time is set between 5ms and 10ms, and the image gain is controlled within the range of 6dB to 12dB to ensure clear images even in low-light environments. Next, parameters are set for the multimodal sensor modules, including flow sensors, water level sensors, and water pressure sensors. Based on acquisition constraints, the sensor sampling frequency is set to 5Hz or 10Hz, and the data output accuracy is standardized; for example, flow parameters are retained to two decimal places, and water pressure data is retained to three decimal places. For the positioning module, the positioning update frequency is set according to the detection path length. Generally, the position update interval is set to 0.2s to 0.5s. At the same time, the inertial navigation parameters are calibrated to keep the positioning error within ±0.2m.

[0030] Continuous video capture of the internal environment of drainage pipes is achieved using an industrial camera mounted on the front of the robot. The industrial camera typically employs a high-resolution CMOS image sensor and is equipped with a wide-angle lens to expand the field of view, allowing simultaneous recording of the pipe's inner wall and bottom areas. During acquisition, the video capture resolution is set to 1920×1080 according to a standardized configuration, with a frame rate maintained at 25fps. An automatic exposure control mechanism dynamically adjusts the exposure time based on changes in lighting conditions inside the pipe. For example, the exposure time is automatically adjusted to 8ms–10ms in darker environments and to around 5ms in brighter areas to maintain stable image brightness. Furthermore, the industrial camera is used in conjunction with a ring-shaped LED supplementary lighting device, with the brightness generally controlled within the range of 300lx–500lx to clearly reveal cracks, deposits, or blockages on the pipe's inner wall. During video acquisition, continuous video data is compressed using H.264 or H.265 encoding to reduce data storage space. Simultaneously, a key image frame is extracted every second, and the corresponding timestamp information is recorded for subsequent defect identification and rapid retrieval. For example, when inspecting a pipe about 40m long, the video acquisition time is about 200s to 300s, which can generate about 5000 to 7500 frames of image data, and at the same time generate about 200 to 300 keyframe images.

[0031] Multimodal sensors mounted on a robot monitor the water flow within drainage pipes to obtain key flow parameters such as flow rate, water level, and water pressure. These sensors typically include electromagnetic flow meters, water level sensors, and pressure sensors, which connect to the robot's control unit via a data acquisition interface. During detection, the flow sensor records drainage flow data at a sampling frequency of 5Hz, with a measurement range generally from 0 m³ / h to 50 m³ / h and an accuracy controllable within ±1%. The water level sensor obtains the water level height within the pipe using ultrasonic or pressure measurement methods, with a sampling frequency typically set to 5Hz to 10Hz, a measurement range of approximately 0m to 1.5m, and an accuracy of approximately ±0.01m. The pressure sensor monitors pressure changes within the pipe, with a measurement range generally from 0MPa to 0.5MPa, a data sampling frequency of approximately 10Hz, and converts the pressure data into a digital signal output. During data acquisition, each sensor records data in a uniform chronological order, and a corresponding timestamp is added to each record. For example, during a 200-second detection process, approximately 1,000 flow rate data records, 1,000 water level data records, and 2,000 water pressure data records can be generated.

[0032] During robot operation, the positioning module collects the robot's position coordinates in the pipeline in real time. The positioning module typically consists of an inertial measurement unit (IMU), an odometry unit, and auxiliary positioning devices, calculating the robot's current position through multi-source information fusion. During inspection, the odometry unit calculates the forward displacement based on the robot's wheeled movement distance, while the IMU records the robot's attitude changes using accelerometers and gyroscopes. These data are then fused using algorithms to obtain continuous spatial position data. The positioning module's update frequency is typically set to approximately 5Hz, meaning it records position coordinate information every 0.2 seconds. Each record includes a timestamp and three-dimensional coordinate data. For example, the position record format can be represented as (t, x, y, z), where t represents the acquisition time, and x, y, and z represent the robot's spatial coordinates within the pipeline. When the robot travels to inspection wells or areas with changes in pipe diameter, the positioning data can be corrected using known structural locations, thereby reducing accumulated errors and keeping the overall positioning error within ±0.2m. When inspecting a drainage pipe approximately 40m long, about 200 spatial position data points can be recorded, forming a complete motion trajectory.

[0033] All collected data is uniformly stored and managed through a mobile terminal configured inside the robot, thus constructing a complete detection dataset. The mobile terminal typically uses embedded computing devices or industrial tablets equipped with large-capacity storage modules, such as 128GB or 256GB solid-state storage, to store the data generated during the detection process. During data storage, video data, sensor data, and positioning data are categorized and recorded according to a unified data structure. Video data is stored in MP4 format and named according to the detection segment number and time sequence, for example, using "segment number + timestamp" for file names. Sensor data is saved in structured data files, such as CSV or JSON formats, recording flow rate, water level, and water pressure parameters. Each record includes fields such as timestamp, sensor number, and sampled value. Positioning data is also recorded in time-series format, for example, using a structure of "timestamp—X coordinate—Y coordinate—Z coordinate". Video keyframes, sensor data, and positioning data are associated in chronological order to establish a unified data index structure, ensuring that each video segment corresponds to specific water flow parameters and spatial location. For example, during the inspection of a 40m drainage pipe, approximately 1GB of video data, thousands of sensor records, and hundreds of location data records can be generated.

[0034] In this embodiment, the adaptive optimization specifically refers to: The video of the inside of the pipeline is subjected to image denoising and contrast enhancement. Perform signal smoothing and outlier identification on water flow parameters, and mark outliers; Remove the outliers; Perform positioning drift analysis on real-time spatial location and extract positioning error parameters; Offset compensation is performed based on the positioning error parameters.

[0035] In this embodiment, after acquiring video data from inside the pipeline, image preprocessing is required to improve image clarity and enhance details within the pipeline. Due to the complex lighting conditions inside the pipeline, and the presence of water mist, reflected light, and equipment vibration, the video footage often exhibits noise, uneven brightness, and insufficient contrast. Continuous video data is split into image frames in chronological order; for example, at a video frame rate of 25fps, 25 frames can be obtained per second. Denoising is then performed on each frame using spatial filtering methods, such as median filtering or Gaussian filtering for image smoothing. Median filtering typically uses a 3×3 or 5×5 neighborhood window, replacing the center pixel value with the median of neighboring pixels to effectively remove salt-and-pepper noise. Gaussian filtering, on the other hand, smooths the image using a Gaussian kernel with a standard deviation σ of approximately 0.8–1.2 to reduce the impact of random noise on image quality. After image denoising, image contrast enhancement is necessary to improve the identifiability of cracks, deposits, or blockages on the pipeline wall. Contrast enhancement can be achieved through histogram equalization, which redistributes the image's grayscale distribution to ensure a uniform distribution of pixel grayscale values ​​within the range of 0 to 255, thereby enhancing the image's brightness levels. Furthermore, in areas with significant localized lighting variations, adaptive contrast enhancement methods can be employed, adjusting image brightness on a localized window basis, such as using an 8×8 pixel window for localized enhancement.

[0036] Because sensors may be affected by factors such as water flow disturbances, equipment vibrations, or electromagnetic interference during operation, the collected data often exhibits short-term fluctuations or abrupt changes, thus requiring signal smoothing. In practice, a moving average method can be used to smooth the raw data. For example, a moving average window of 5-10 sampling points can be set, and the data within the window can be averaged to obtain a smoothed water flow parameter sequence. For instance, when the sensor sampling frequency is 5Hz, a window length of 10 sampling points corresponds to a data interval of approximately 2 seconds. This method can effectively reduce the impact of short-period fluctuations on data stability. Besides the moving average method, an exponentially weighted smoothing method can also be used. By setting a smoothing coefficient α between 0.3 and 0.5, current data is weighted and fused with historical data to obtain a more stable water flow parameter variation curve. After signal smoothing, outlier identification is necessary. Outlier identification can be achieved through statistical analysis methods, such as calculating the average and standard deviation of data over a period of time. When a data point deviates from the average by more than three times the standard deviation, it can be identified as an outlier. Simultaneously, reasonable ranges can be set for the project. For example, the water level in drainage pipes is usually between 0m and 1.2m, and the water pressure is generally between 0MPa and 0.3MPa. When the data exceeds these ranges, it can be identified as an anomaly. For the identified abnormal data, it is necessary to mark it in the data record, for example, by adding an anomaly identifier field, so that it can be distinguished in subsequent processing.

[0037] The data sequence is filtered based on the anomaly identification field, and data points marked as anomalous are extracted. Isolated outliers can be directly deleted, i.e., removed from the data sequence. For example, if a water pressure suddenly jumps from 0.15 MPa to 0.45 MPa, while the preceding and following data are all within the range of 0.14 MPa to 0.16 MPa, this point can be considered an anomaly and deleted. After deleting outliers, the data sequence needs to be compensated to maintain the continuity of the data time series. Compensation methods typically employ linear interpolation or nearest-neighbor average substitution. For example, after deleting an outlier data point in the time series, a new data value can be obtained by linear interpolation based on the two preceding and following normal data points, thus filling the missing position. For instance, if the water level at one moment is 0.42 m and the water level at the next moment is 0.44 m, the missing position can be compensated to 0.43 m. Furthermore, when two or more consecutive outlier data points appear, the average value of the nearest interval can be used as a substitute to avoid excessive interpolation errors.

[0038] The spatial location data collected by the robot is arranged chronologically, with each record containing a timestamp and corresponding spatial coordinates, such as (t, x, y, z). The positioning trajectory is compared and analyzed by combining pipeline topology information and known structural locations, such as manhole coordinates or pipe node locations. When the robot passes a manhole or other known fixed location, the actual coordinates of that location are compared with the coordinates recorded by the robot to calculate the positioning error. For example, when the actual coordinates of the manhole are (X0, Y0, Z0), while the robot records the location as (X1, Y1, Z1), the error value can be obtained through three-dimensional distance calculation. Statistical analysis of multiple correction points reveals the trend of positioning error over time. For example, when inspecting a pipe approximately 40m long, the positioning error may gradually accumulate from 0.05m at the starting point to around 0.25m. Statistical analysis of this error data allows for the extraction of positioning error parameters, such as the average error value, the maximum error value, and the error growth rate.

[0039] An error compensation model is established based on the positioning drift analysis results. For example, a linear compensation relationship can be constructed based on the law of error variation with distance. When the error growth rate is approximately 0.005 m / m, the corresponding error compensation value can be calculated based on the robot's travel distance, and this compensation value is corrected from the original position coordinates. For example, when the robot travels 20 m, the cumulative error is approximately 0.1 m, which can be corrected by subtracting the corresponding offset from the positioning coordinates. For cases with multiple correction points, a segmented compensation method can be used, that is, the positioning error is distributed and corrected between two correction points according to a linear ratio, making the positioning trajectory closer to the actual path. In addition, the corrected positioning data can be further processed by trajectory smoothing methods, such as using three-point smoothing or moving average to reduce position jitter and make the trajectory more continuous and stable. After completing the offset compensation, a new spatial position data sequence is obtained, where each record contains corrected three-dimensional coordinate information.

[0040] In this embodiment, the specific steps of the indexation analysis are as follows: Visual recognition based on video footage inside the pipeline is used to mark the location of blockages inside the pipeline. Based on the location of siltation inside the pipeline, the siltation depth and cross-sectional blockage rate are calculated to obtain siltation assessment indicators; Perform pipe structure feature detection on the video of the inside of the pipe and mark multiple structural feature points; Structural damage is analyzed based on multiple structural feature points, and the damage level is output. Based on the water flow parameters, a quantitative analysis of operational smoothness is performed to obtain the water flow smoothness coefficient.

[0041] In this embodiment, the processed continuous video is split into image frames in chronological order. For example, at a video frame rate of 25fps, 25 frames of image data can be acquired per second. Feature extraction is performed on each frame, and image segmentation and texture recognition methods are used to identify the sediment areas at the bottom of the pipe. Normal areas within the pipe typically have a relatively uniform grayscale distribution, while sediment areas often exhibit obvious color differences and texture variations. Therefore, grayscale thresholding or region growing methods can be used for preliminary image segmentation. For example, areas with grayscale values ​​below a certain threshold (e.g., grayscale values ​​less than 60) are identified as areas where sediment may exist. Edge detection methods are used to extract the contours of these areas, such as using gradient edge detection to identify the boundary morphology of the sediment areas. After identifying suspected sediment areas, consistency judgment is performed using consecutive frame images. When a location exhibits similar sediment features in 5 to 10 consecutive frames, the area can be identified as a stable blockage area, and its temporal position and corresponding spatial coordinates in the video are recorded. In this way, sediment or blockage areas can be marked in the video. For example, during the inspection of a drainage pipe approximately 40m long, 2 to 4 obvious siltation areas may be identified, each area being approximately 0.5m to 2m in length.

[0042] The pipe cross-sectional profile is determined based on the pipe structural features in the video image. For example, the inner wall profile of the pipe can be identified using a circular edge detection method, and the pipe cross-sectional diameter can be calculated. Since drainage pipes are typically circular, a circular profile can be fitted based on the pipe edge pixels in the image to obtain the pipe cross-sectional radius. For example, in a DN400 pipe, the theoretical diameter is approximately 400 mm, and the ratio between pixels and the actual size can be calculated using image scaling. The height of the deposition area in the image is measured, i.e., the vertical distance between the top of the deposit and the bottom of the pipe is calculated, to obtain the sedimentation depth. For example, in a certain detection area, the sediment height is approximately 80 mm, and the pipe radius is approximately 200 mm, then the sedimentation depth in this area accounts for approximately 40% of the pipe radius. Further calculation of the cross-sectional blockage rate is required, i.e., the proportion of the pipe cross-sectional area occupied by the sediment. Specifically, this is calculated based on the ratio between the area of ​​the deposition area in the image and the total cross-sectional area of ​​the pipe. For example, when the area of ​​the deposition area accounts for approximately 30% of the cross-sectional area, the cross-sectional blockage rate for that area is 0.30. By combining the two indicators of siltation depth and cross-sectional blockage rate, a comprehensive siltation assessment index can be formed. For example, areas with a cross-sectional blockage rate of less than 20% are judged as mild siltation, those between 20% and 50% are judged as moderate siltation, and areas with a blockage rate of more than 50% are judged as severe blockage.

[0043] The contour of the pipe's inner wall is extracted from the video image, and edge detection methods are used to identify the pipe's inner wall boundaries, thereby determining the pipe's structural contour. After obtaining the contour information, key locations in the image are identified using feature point extraction methods, such as crack initiation points, crack end points, interface connections, and pipe deformation locations. These structural feature points typically appear as areas with significant grayscale changes or abrupt morphological changes, and can therefore be identified using corner detection or curvature analysis methods. For example, at pipe joint locations, obvious straight edge structures usually appear, while crack areas may appear as thin, dark lines. By analyzing these structural features in the image, multiple structural feature points can be marked. For example, during the inspection of a 40m long pipe, 10 to 20 key structural feature points may be detected, including interface locations, crack locations, and deformation areas. Each feature point needs to have its temporal location and corresponding spatial coordinates recorded in the video, and a feature point database is established. First, each feature point is categorized, for example, based on morphological characteristics into crack feature points, interface misalignment feature points, and pipe wall deformation feature points. A comprehensive evaluation is then performed based on the number, distribution, and size of the feature points. For example, crack length can be calculated by converting the image pixel length to the actual size ratio; a crack length exceeding 50mm is considered a significant crack, while a length exceeding 150mm may affect the stability of the pipe structure. Crack width also needs to be evaluated; for example, a crack width greater than 3mm is generally considered to indicate severe structural damage. For interface misalignment, the difference in position between the two edges of the interface can be analyzed; a misalignment height exceeding 10mm indicates significant deformation of the interface connection. Statistical analysis of all structural feature points yields the degree of pipe structural damage. For example, a small number of short cracks indicates minor damage; a large number of cracks, with some exceeding 100mm in length, indicates moderate damage; and large-scale cracks or significant structural deformation indicate severe damage.

[0044] A comprehensive analysis of parameters such as flow rate, water level, and water pressure is conducted, and the changes in water flow within the pipeline are observed through time-series data. Under normal circumstances, the water flow in a drainage pipeline should maintain a stable flow state, with a certain proportional relationship between flow rate and water level. For example, in a DN400 drainage pipeline, when the water level is maintained within the range of 0.3m to 0.5m, the flow rate is usually between 10m³ / h and 20m³ / h. When siltation or blockage occurs inside the pipeline, the water flow resistance increases, leading to a rise in water level and a decrease in flow rate. Therefore, the pipeline's unobstructed flow can be assessed by analyzing the relationship between water level and flow rate. The specific quantification method is to calculate the proportional relationship between the current flow rate and the design flow rate. For example, if the design flow rate is 20m³ / h, and the actual monitored flow rate is 14m³ / h, then the flow utilization rate is 0.70. Simultaneously, a flow rate index is calculated based on water level changes; for example, when the water level is consistently higher than the upper limit of the normal range, it indicates a decrease in drainage capacity. By comprehensively calculating the flow utilization rate and the water level change rate, a water flow coefficient can be obtained, which typically ranges from 0 to 1. For example, a flow rate coefficient greater than 0.8 indicates that the drainage pipe is operating well; a coefficient between 0.5 and 0.8 indicates a certain degree of blockage; and a coefficient below 0.5 indicates a significant decrease in drainage capacity, requiring dredging.

[0045] In this embodiment, the specific steps of step S5 are as follows: Based on multidimensional analysis indicators, time distribution matching is performed to output indicator distribution information at different locations; Based on the distribution information of the aforementioned indicators, a congestion risk analysis is performed to extract the congestion risk level. Based on the blockage risk level, the pipeline distribution is fitted to obtain a blockage risk distribution map; Develop dredging operation routes based on the blockage risk distribution map and output the operation routes; A comprehensive report is generated based on the blockage risk distribution map and operation path; the comprehensive report is then uploaded to the cloud-based dredging management platform to complete the data collection and analysis.

[0046] In this embodiment, timestamps are used as a unified index to synchronously match video recognition results, water flow parameter data, and positioning data. Since video keyframes are typically extracted at 1-second intervals, while sensor data sampling intervals are usually 0.2-0.5 seconds, the keyframe time is used as a reference time axis to match the sensor data and positioning data closest to that time point. For example, when the keyframe time is t1, data with a time difference less than 0.3 seconds can be selected as the corresponding data record, thus achieving time alignment between different data sources. After time matching, each record is associated with its corresponding spatial location to obtain the siltation assessment indicators, structural damage level, and water flow coefficient corresponding to a specific location in the pipeline. For example, in a drainage pipe approximately 40 meters long, the space can be divided into 0.5-meter or 1-meter units, forming 40-80 data sampling locations. For each location unit, the corresponding siltation depth, cross-sectional blockage rate, structural damage level, and water flow coefficient are recorded. This method establishes a database of indicator distribution for each location within the pipeline, forming multi-dimensional spatial distribution information of indicators. For example, the blockage rate of a section at one location is 0.35, the siltation depth is 90mm, and the water flow coefficient is 0.65, while the blockage rate at another location is only 0.10 and the flow coefficient reaches 0.85.

[0047] A blockage risk assessment model is established, using indicators such as siltation depth, cross-sectional blockage rate, water flow coefficient, and structural damage level as input parameters. Weighting coefficients are assigned based on the impact of each indicator on drainage capacity; for example, the weight for cross-sectional blockage rate can be set to 0.4, for siltation depth to 0.25, for water flow coefficient to 0.2, and for structural damage level to 0.15. These weights are used to comprehensively calculate the risk index for each spatial location. For instance, in a certain pipe section, when the cross-sectional blockage rate is 0.40, the siltation depth is 100mm, and the water flow coefficient is 0.55, the comprehensive risk index may reach above 0.65. Based on the range of the risk index, blockage risk can be divided into multiple levels; for example, a risk index below 0.30 is considered a low-risk area, 0.30–0.60 is considered a medium-risk area, and above 0.60 is considered a high-risk area. In addition, cluster analysis can be performed on multiple consecutive high-risk locations. When the length of a continuous area exceeds 3m to 5m, it indicates that there is a significant risk of blockage in the area, and dredging operations should be prioritized.

[0048] Based on the pipeline network topology information, each risk data point is matched with its corresponding pipe segment location and marked in the pipeline network structure model. Risk levels are smoothed using spatial interpolation or curve fitting methods to make the risk distribution more continuous. For example, risk nodes can be established in units of 0.5m or 1m, and interpolation calculations can be performed based on the risk index between adjacent nodes to form a continuous risk curve. In this way, originally discrete risk points can be expanded into continuous risk areas. In the pipeline network structure diagram, different risk levels can be distinguished using different marking methods; for example, low-risk areas can be marked in green, medium-risk areas in yellow, and high-risk areas in red. Taking a drainage pipe of approximately 100m in length as an example, there may be a medium-risk area within a 20m range and a high-risk area within a 5m range.

[0049] Based on the pipeline topology, all accessible manholes or inlets are identified, and these locations serve as the starting or ending points for work path planning. Pipe sections are prioritized according to their blockage risk level; for example, high-risk areas have the highest priority, followed by medium-risk areas, while low-risk areas can be temporarily deferred. Multiple high-risk areas are connected using path planning methods to form the optimal work route. For instance, when two high-risk areas are close together, they can be planned along the same work path to reduce redundant equipment movement. The path planning process also needs to consider the movement speed and operation time of the robot or dredging equipment. For example, when the dredging equipment moves at approximately 0.2 m / s, clearing a 10 m long severely blocked area may take about 10–15 minutes. By comprehensively considering risk level, spatial location, and operational efficiency, one or more optimal dredging work paths can be generated. For example, in a drainage network, two main work paths are planned, each approximately 30–50 m long, covering all high-risk areas.

[0050] The video data, water flow parameter data, structural inspection results, and blockage risk analysis results obtained during pipeline inspection are uniformly organized and categorized according to pipe segment number. A blockage risk distribution map and a planned dredging operation route map are included in the report, enabling a clear visual representation of the drainage network's operational status. For example, the report can list key indicators such as siltation depth, cross-sectional blockage rate, structural damage level, and water flow smoothness coefficient for each pipe segment, along with their corresponding spatial locations. The report also provides risk level statistics, such as a high-risk area of ​​12m, a medium-risk area of ​​25m, and other areas classified as low-risk. The planned dredging operation route information is added to the report, along with recommended operation sequence and estimated operation time, such as an estimated 2-3 hours for completing all dredging operations. After the report is generated, it is uploaded to the cloud-based dredging management platform via a network communication interface and stored in the platform's database. Upon receiving the report, the platform allows operation and maintenance personnel to view and make decisions, thus completing the data collection and analysis loop for the tunnel drainage system dredging project.

[0051] In this embodiment, the specific steps for performing time distribution matching based on multidimensional analysis indicators and outputting indicator distribution information at different locations are as follows: Calculate the real-time spatial location timestamp; Data processing time is calculated for the multidimensional analysis indicators to obtain the timestamps of the analysis indicators; Based on the location timestamp, the timestamps of the analysis indicators are matched for time distribution, and the indicator distribution information at different locations is output.

[0052] In this embodiment, a unified time reference is established when the robot starts its inspection operation. For example, the internal clock of the mobile terminal is used as the reference time source, and a time synchronization mechanism is used to uniformly calibrate each acquisition module, ensuring that the industrial camera, multimodal sensor, and positioning module all record data using the same time reference. When the positioning module outputs spatial position data, a time stamp is added to each data record. For example, when the positioning module's update frequency is set to 5Hz, that is, a new position data record is generated every 0.2s, each record contains a timestamp and corresponding spatial coordinate data. The data structure can be represented as (t, x, y, z), where t is the positioning timestamp, and x, y, and z are the robot's three-dimensional coordinates in the pipe. When inspecting a drainage pipe approximately 40m long, the robot's movement speed is typically controlled at around 0.2m / s, and the entire inspection process lasts approximately 200s, thus recording approximately 1000 positioning data records. To further improve the accuracy of time recording, the timestamp is typically recorded in millisecond-level time units, for example, in the format of "second.millisecond". This method can accurately describe the changes in the robot's trajectory in the pipeline, so that each spatial position can correspond to an accurate point in time, thus providing basic data for subsequent data time matching.

[0053] Multidimensional analysis indicators mainly include data such as siltation depth, cross-sectional blockage rate, structural damage level, and water flow smoothness coefficient. These indicators are usually calculated from video image analysis and water flow parameter analysis. In the specific implementation process, video data and sensor data are processed. When a video frame completes image recognition and calculates the corresponding siltation assessment indicator, the generation time of that indicator needs to be recorded. For example, video keyframes are usually extracted at 1-second intervals. Therefore, after each frame image is recognized, a set of analysis results is generated. For example, the siltation depth corresponding to this frame image is 85mm, the cross-sectional blockage rate is 0.32, and the structural feature information of the area can also be obtained. While generating these indicator data, a data processing time identifier, i.e., a timestamp of the analysis indicator, is added to each record. For example, when a frame image is processed at time t1, this time is recorded as the timestamp of the corresponding indicator. For water flow parameter analysis data, its processing time also needs to be recorded. For example, when the sensor sampling frequency is 5Hz, a set of water flow parameter data is generated every 0.2s. After smoothing and outlier removal of these data, a time identifier also needs to be added to the processing results. In this way, complete time-series data of analytical indicators can be generated. For example, in a 200-second detection process, approximately 200 video analysis indicator records and approximately 1,000 water flow parameter analysis records can be generated, each record containing a corresponding timestamp.

[0054] The positioning time series is used as the spatial reference time axis because the positioning data directly reflects the robot's spatial position changes in the pipeline. For example, the positioning module records position data every 0.2 seconds, while video analysis metrics typically generate a record every 1 second. Therefore, a time matching method is needed to map the two types of data. The positioning time series is searched for the positioning time point closest to the timestamp of the analysis metric. For example, when the timestamp of a certain analysis metric is t2, the data with the smallest time difference (less than 0.3 seconds) is selected from the positioning data as the corresponding position record. This method determines the spatial coordinates of the analysis metric. For example, if the siltation depth data generated at a certain time point is 90mm, its timestamp is t2. By matching, the corresponding position coordinates are (x2, y2, z2), and the metric can be marked at that spatial location. After completing all data matching, the pipeline space can be divided at certain intervals, such as every 1m or 0.5m, into spatial units, and the metric data in each spatial unit can be statistically analyzed. For example, in a 40m long pipe, 40 location nodes can be obtained by dividing the pipe into 1m sections. Each node can record information such as the corresponding siltation depth, cross-sectional blockage rate, and water flow coefficient. Through this time distribution matching method, multidimensional analysis indicators can be mapped to specific spatial locations, thus forming complete indicator distribution information. For example, one area may have a siltation depth of 80mm and a flow coefficient of 0.60, while another location may only have slight siltation.

[0055] In this embodiment, an intelligent data acquisition and processing system for dredging projects in a tunnel drainage system is provided. In this system, a data acquisition and setting module receives work tasks from a cloud-based dredging management platform and generates a data acquisition constraint set, which is then input as work configuration to a work detection module. The latter drives a pipeline detection robot to collect video, sensor, and positioning data to form a first dataset. A data optimization module performs noise reduction, smoothing, anomaly removal, and positioning compensation on the first dataset to generate a high-quality second dataset. An analysis module performs index processing on the second dataset, extracting multi-dimensional analysis indicators such as siltation depth, cross-sectional blockage rate, structural damage level, and water flow coefficient. A planning module plans the dredging operation path based on the multi-dimensional analysis indicators and synchronously uploads the results to the cloud platform, realizing data-driven intelligent dredging operations. The method for executing the aforementioned intelligent data acquisition and processing method for dredging projects in a tunnel drainage system includes: The data acquisition and setting module is used to receive and process tasks based on the cloud-based dredging and management platform; and to set data acquisition constraints based on the processing tasks to obtain a set of data acquisition constraints. The task detection module is used to drive the pipeline detection robot to perform task detection based on the collected constraint set and output the first dataset. The data optimization module is used to adaptively optimize the first dataset and output the second dataset. The analysis module is used to perform index-based analysis on the second dataset and output multidimensional analysis indicators. The planning module is used to plan dredging operation paths based on multi-dimensional analysis indicators and upload them synchronously to the cloud-based dredging management platform to complete the data collection and analysis.

[0056] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0057] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligent data acquisition and processing of dredging projects in tunnel drainage systems, characterized in that, Includes the following steps: Step S1: Receive and process tasks based on the cloud-based dredging management platform; Based on the processing task, data acquisition constraints are set to obtain a data acquisition constraint set. Step S2: Drive the pipeline detection robot to perform operational detection based on the collected constraint set, and output the first dataset; Step S3: Perform adaptive optimization on the first dataset and output the second dataset; Step S4: Perform index-based analysis on the second dataset and output multidimensional analysis indicators; Step S5: Plan the dredging operation path based on multi-dimensional analysis indicators, and upload it to the cloud-based dredging management platform to complete the data collection and analysis.

2. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 1, characterized in that, The specific steps of step S1 are as follows: Tasks are received and processed based on a cloud-based dredging management platform; Based on the processing task, relevant design documents are extracted; the relevant design documents are analyzed for pipeline topology to extract pipe segment connection relationships, node positions and pipe diameter parameters, and pipeline topology information is generated; Based on the relevant design documents, information analysis was performed to obtain basic pipeline information; Based on the pipeline basic information and pipeline topology information, data acquisition constraints are set to obtain a data acquisition constraint set; the data acquisition constraint set includes video resolution, sensor sampling frequency, positioning module accuracy, and data sampling interval.

3. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 2, characterized in that, The relevant design documents include tunnel name, section, drainage pipeline number, manhole location, pipe diameter parameters, historical maintenance records, and design drawings.

4. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 1, characterized in that, The specific steps of step S2 are as follows: Standardized operation configurations are set based on the collection constraint set; The pipeline detection robot is driven to perform detection operations and extract the detection dataset based on standardized operating configurations. Perform data format conversion and timeline synchronization alignment on the probe dataset to output the first dataset.

5. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 4, characterized in that, The operation detection specifically includes; The pipeline inspection robot device includes an industrial camera, a multimodal sensor, and a mobile terminal; The configuration of the pipeline inspection robot is adjusted according to the standardized operation configuration described above; Using industrial cameras to capture video of the inside of pipes; Water flow parameters are obtained by collecting flow rate, water level, and water pressure in the pipeline using multimodal sensors. Real-time spatial location is collected based on the positioning module; The system uses mobile terminals to locally store videos of the pipeline interior, water flow parameters, and real-time spatial location, thus constructing a detection dataset.

6. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 1, characterized in that, The adaptive optimization specifically refers to: The video of the inside of the pipeline is subjected to image denoising and contrast enhancement. Perform signal smoothing and outlier identification on water flow parameters, and mark outliers; Remove the outliers; Perform positioning drift analysis on real-time spatial location and extract positioning error parameters; Offset compensation is performed based on the positioning error parameters.

7. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 1, characterized in that, The specific steps of the index-based analysis are as follows: Visual recognition based on video footage inside the pipeline is used to mark the location of blockages inside the pipeline. Based on the location of siltation inside the pipeline, the siltation depth and cross-sectional blockage rate are calculated to obtain siltation assessment indicators; Perform pipe structure feature detection on the video of the inside of the pipe and mark multiple structural feature points; Structural damage is analyzed based on multiple structural feature points, and the damage level is output. Based on the water flow parameters, a quantitative analysis of operational smoothness is performed to obtain the water flow smoothness coefficient.

8. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 1, characterized in that, The specific steps of step S5 are as follows: Based on multidimensional analysis indicators, time distribution matching is performed to output indicator distribution information at different locations; Based on the distribution information of the aforementioned indicators, a congestion risk analysis is performed to extract the congestion risk level. Based on the blockage risk level, the pipeline distribution is fitted to obtain a blockage risk distribution map; Develop dredging operation routes based on the blockage risk distribution map and output the operation routes; A comprehensive report is generated based on the blockage risk distribution map and operation path; the comprehensive report is then uploaded to the cloud-based dredging management platform to complete the data collection and analysis.

9. The intelligent data acquisition and processing method for dredging projects in tunnel drainage systems according to claim 8, characterized in that, The specific steps for matching the time distribution based on multidimensional analysis indicators and outputting the indicator distribution information at different locations are as follows: Calculate the real-time spatial location timestamp; Data processing time is calculated for the multidimensional analysis indicators to obtain the timestamps of the analysis indicators; Based on the location timestamp, the timestamps of the analysis indicators are matched for time distribution, and the indicator distribution information at different locations is output.

10. A smart data acquisition and processing system for dredging projects in tunnel drainage systems, characterized in that, The method for intelligent data acquisition and processing of dredging engineering data for a tunnel drainage system as described in claim 1 includes: The data acquisition and setting module is used to receive and process tasks based on the cloud-based dredging and management platform; and to set data acquisition constraints based on the processing tasks to obtain a set of data acquisition constraints. The task detection module is used to drive the pipeline detection robot to perform task detection based on the collected constraint set and output the first dataset. The data optimization module is used to adaptively optimize the first dataset and output the second dataset. The analysis module is used to perform index-based analysis on the second dataset and output multidimensional analysis indicators. The planning module is used to plan dredging operation paths based on multi-dimensional analysis indicators and upload them synchronously to the cloud-based dredging management platform to complete the data collection and analysis.

Images