Intelligent detection and evaluation method and system for underground pipeline defects

Abstract

The invention relates to an intelligent detection and evaluation method and system for underground pipeline defects, and the method comprises the following steps: S1, collecting multi-modal data of an underground pipeline based on an inspection robot, and transmitting the multi-modal data to an intelligent detection and evaluation platform; s2, the intelligent detection and evaluation platform preprocesses the collected multi-modal data; s3, performing defect detection on the underground pipeline based on the preprocessed multi-modal data; s4, constructing a three-dimensional model of the underground pipeline, and superposing a defect detection result into the three-dimensional model; and S5, based on the three-dimensional model of the underground pipeline, performing intelligent evaluation on the underground pipeline, and generating a pipeline condition evaluation report. According to the invention, the efficiency and reliability of underground pipeline defect detection are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of pipeline detection, and in particular to an intelligent detection and evaluation method and system for underground pipeline defects. Background Art

[0002] With the acceleration of urbanization and the aging of infrastructure, underground pipelines, as the carriers of key resources such as urban water, electricity, gas and communications, are directly related to the normal operation of the city. However, underground pipelines that have been in operation for a long time are easily affected by factors such as the external environment, material aging, corrosion, and accumulation of sediments in the pipes, resulting in defects such as cracks, corrosion, blockage, deformation and even rupture. These problems may not only lead to economic losses such as waste of resources and road collapse, but also threaten the safety and stability of the city. Traditional pipeline inspection methods mostly rely on manual visual inspection and simple tool testing, which are not only inefficient and have limited effects, but are also easily interfered by human factors, resulting in missed detection or misjudgment of defects. Summary of the invention

[0003] In order to solve the above problems, the purpose of the present invention is to provide an intelligent detection and evaluation method and system for underground pipeline defects, which effectively improves the efficiency and reliability of underground pipeline defect detection.

[0004] To achieve the above object, the present invention adopts the following technical solutions: An intelligent detection and evaluation method for underground pipeline defects comprises the following steps: S1: Based on the inspection robot, collect multimodal data of underground pipelines and transmit them to the intelligent detection and evaluation platform; S2: Intelligent detection and evaluation platform, pre-processing the collected multi-modal data; S3: Based on the pre-processed multimodal data, defect detection of underground pipelines is performed, including: Image recognition algorithms based on convolutional neural networks detect visible defects, including cracks, corrosion, and blockages; Based on the RANSAC point cloud data processing algorithm, the morphological changes of the pipeline are detected and the deformation area is located; Ultrasonic signal defect detection based on Fourier transform and time-frequency analysis; S4: constructing a three-dimensional model of the underground pipeline and superimposing the defect detection results on the three-dimensional model; S5: Based on the 3D model of underground pipelines, intelligently evaluate the underground pipelines and generate a pipeline condition assessment report.

[0005] Furthermore, multimodal data includes: image data of the inside of the pipeline obtained by a high-definition camera; three-dimensional point cloud data of the inside of the pipeline constructed by a lidar; depth detection signals for pipe wall thickness detection or potential defect detection obtained by an ultrasonic sensor; humidity and temperature obtained by an environmental sensor; and positioning information of the inspection robot recorded by an IMU and an odometer.

[0006] Further, preprocessing is as follows: S21: Remove duplicate, ambiguous or redundant sampling data from multimodal data; S22: Process different types of data separately, including: Image data, use Gaussian filtering to eliminate image noise, enhance pipeline feature details, apply histogram equalization to improve image contrast, and use the camera calibration matrix to eliminate image distortion caused by high-definition cameras; For point cloud data, use the neighborhood filtering method to remove invalid point cloud noise and fill in the missing parts through edge fitting or interpolation; The deep detection signal is de-noised using a Butterworth low-pass filter, and the envelope signal is extracted using the Hilbert transform to capture the waveform anomaly caused by the defect; Humidity and temperature data obtained by environmental sensors, using time series interpolation to fill in missing data and normalize different environmental parameters; S23: Align the multimodal data after being processed separately in time and space to unify the multimodal data format.

[0007] Furthermore, the image recognition algorithm based on convolutional neural network adopts the Mask R-CNN target detection framework, including convolutional feature extraction network, region proposal network, ROI alignment and detection network: The convolutional feature extraction network uses convolutional layers to extract features from images. The feature extraction process is as follows: ; in, is the value of the feature map of the lth layer at position (i, j); For the The layer outputs the value of each position in the feature map; is the weight value of the convolution kernel; a, b are position indexes; is the bias of the feature map of the lth layer; σ is the activation function; Multiple convolutional layers are stacked to form a feature pyramid, where the lower layers extract texture details and the higher layers encode global semantics. Based on the extracted features, a candidate region ROI that may contain defects is generated based on the region proposal network; For the ROI that passes the screening, ROI Align is used to accurately extract the features that match the ROI, and the features are passed to the following branches: Classification branch, target category classification prediction, loss function L cls is the cross entropy loss:

[0008] Among them, K is the total number of categories; y k is the one-hot label corresponding to the true category k; P k is the probability distribution of the predicted category k; The positioning branch accurately locates the target, optimizes the IoU of the target bounding box, improves the overlap between the predicted box and the real bounding box, and reduces the positioning loss. L loc for: ; Among them, B true is the true bounding box of the candidate region; B pred is the predicted target bounding box; The mask branch generates target instance masks, generates a pixel-level mask of a specific category for each ROI, and calculates the mask error using binary cross entropy loss L mask : ; Among them, M i,j is the true mask at position (i, j); Predict the mask at position (i, j); Output the category, location and mask of each candidate object.

[0009] Furthermore, based on the RANSAC point cloud data processing algorithm, the morphological changes of the pipeline are detected, as follows: Using the pre-processed pipeline point cloud data, extract the reference cross-sectional shape of the current pipeline; The plane segmentation algorithm is applied to divide the point cloud into several two-dimensional sections for frame-by-frame analysis of the pipeline cross-sectional morphology; The random sampling consensus algorithm RANSAC is used to fit the shape and fit the pipe section from the point cloud. According to the fitting result, it is judged whether the section deviates from the ideal shape. The deviation formula is: ; Among them, r ideal is the pipe cross section, r actual is the radius of the fitting section. If d exceeds the set threshold (e.g. d>0.05), it is determined to be a deformation area; Mark the area where the pipeline is deformed, give the deformation type, and output the deformation deviation d.

[0010] Further, ultrasonic signal defect detection based on Fourier transform and time-frequency analysis is as follows: For the collected ultrasonic or sonar signal x(t), the signal spectrum characteristics are extracted by fast Fourier transform: ; Where X(f) is the frequency domain signal; is an imaginary unit; f and t represent frequency and time respectively; Analyze the peak position of the echo signal. The echo peak indicates the strong signal generated by the reflection from the material interface. The time interval Δt of different peaks corresponds to the thickness. T : ; where c is the speed of sound in the material; If the time intervals are not uniform, it is considered that there is material delamination or internal cracks; Perform envelope analysis on the echo signal and evaluate the pipe wall strength by the attenuation rate of the signal amplitude with depth:

[0011] Where c is the speed of sound in the material; A0 is the initial signal amplitude; α is the attenuation coefficient; is the propagation distance; if the attenuation A is higher than the threshold, there is a material structure problem; And use short-time Fourier transform (STFT) to decompose the time-frequency characteristics of the signal and locate the specific time point when the pipe wall defect occurs: ; Where w(t) is the window function; is the time-integrated variable; f is the frequency; Locate the defect position based on characteristic frequency and energy changes; Finally, the wall thickness deviation value, delamination or internal crack risk area are output.

[0012] Furthermore, S4 is specifically: Based on the preprocessed point cloud data, the ICP algorithm is used to align the multi-frame point clouds; The Delaunay triangulation algorithm is used to generate a three-dimensional surface mesh, and each fitted section in the point cloud forms a continuous pipe wall surface through an interpolation algorithm; The preprocessed image is then wrapped onto the generated 3D surface mesh to enhance the texture realism and construct a 3D model of the underground pipeline; Align the underground pipeline 3D model with the geographic coordinate data, obtain the time series of the inspection robot's position coordinates from the IMU and odometer, align them according to the pipeline layout reference data in the geographic information system GIS, and convert the robot's reference coordinate system into the GIS global coordinate system; By using Mask R-CNN to detect the defect location in the image and obtain the bounding box or pixel area, the pixel coordinates of the image detection result are converted into 3D model coordinates; and the point cloud abnormal area is directly matched to the 3D model coordinates; the corresponding 3D coordinates of each ultrasonic sampling point are calculated according to the ultrasonic signal defect detection results; All inspection results are aligned to a unified underground pipeline 3D model reference coordinate system and the defect types are marked.

[0013] Furthermore, S5 is specifically: Quantitatively score the defects according to their type and severity to form the pipeline health score PHI; ; in, According to the defect type The severity of the disease is graded and scored; Defect type The weight of; n is the total number of defect types; And evaluate the repair priority based on defect type priority, severity and location distribution; An assessment report is generated based on the test results and model assessment, including basic information, defect distribution summary and comprehensive score PHI; the basic information includes pipeline number, location coordinates, material, length, diameter and service life; the defect distribution summary includes the number, location, type and severity of defects.

[0014] An intelligent detection and evaluation system for underground pipeline defects includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the intelligent detection and evaluation method for underground pipeline defects as described above.

[0015] The present invention has the following beneficial effects: 1. The present invention integrates multimodal data (images, point clouds, ultrasound, etc.), uses deep learning algorithms to detect explicit defects (such as cracks, corrosion, and blockage), combines geometric modeling to evaluate pipeline deformation, and uses frequency domain analysis to identify hidden defects such as changes in pipeline wall thickness. This enables the pipeline health status to be obtained in an all-round and multi-dimensional manner; 2. The present invention uses 3D modeling technology to superimpose defect information into the 3D model of the pipeline, which not only greatly improves the accuracy of defect location and evaluation, but also provides a more scientific and intuitive reference for subsequent pipeline maintenance; 3. The present invention calculates the health score of the pipeline based on industry standards and historical data to achieve full life cycle management of the underground pipeline network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0017] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments: refer to Figure 1 In this embodiment, an intelligent detection and evaluation method for underground pipeline defects is provided, comprising the following steps: S1: Based on the inspection robot, collect multimodal data of underground pipelines and transmit them to the intelligent detection and evaluation platform; S2: Intelligent detection and evaluation platform, pre-processing the collected multi-modal data; S3: Based on the pre-processed multimodal data, defect detection of underground pipelines is performed, including: Image recognition algorithms based on convolutional neural networks detect visible defects, including cracks, corrosion, and blockages; Based on the RANSAC point cloud data processing algorithm, the morphological changes of the pipeline are detected and the deformation area is located; Ultrasonic signal defect detection based on Fourier transform and time-frequency analysis; S4: constructing a three-dimensional model of the underground pipeline and superimposing the defect detection results on the three-dimensional model; S5: Based on the 3D model of underground pipelines, intelligently evaluate the underground pipelines and generate a pipeline condition assessment report.

[0018] In this embodiment, the multimodal data includes: image data of the inside of the pipeline obtained by a high-definition camera; three-dimensional point cloud data of the inside of the pipeline constructed by a lidar; depth detection signals for pipe wall thickness detection or potential defect detection obtained by an ultrasonic sensor; humidity and temperature obtained by an environmental sensor; and positioning information of the inspection robot recorded by an IMU (inertial measurement unit) and an odometer.

[0019] In this embodiment, the preprocessing is as follows: S21: Remove duplicate, ambiguous or redundant sampling data from multimodal data; S22: Process different types of data separately, including: Image data, use Gaussian filtering to eliminate image noise, enhance pipeline feature details, apply histogram equalization to improve image contrast, and use the camera calibration matrix to eliminate image distortion caused by high-definition cameras; For point cloud data, use the neighborhood filtering method to remove invalid point cloud noise and fill in the missing parts through edge fitting or interpolation; The deep detection signal is de-noised using a Butterworth low-pass filter, and the envelope signal is extracted using the Hilbert transform to capture the waveform anomaly caused by the defect; Humidity and temperature data obtained by environmental sensors, using time series interpolation to fill in missing data and normalize different environmental parameters; S23: Align the multimodal data after being processed separately in time and space to unify the multimodal data format.

[0020] In this embodiment, the image recognition algorithm based on convolutional neural network adopts the Mask R-CNN target detection framework, including convolutional feature extraction network, region proposal network, ROI alignment and detection network: The convolutional feature extraction network uses convolutional layers to extract features from images. The feature extraction process is as follows: ; in, is the value of the feature map of the lth layer at position (i, j); For the The layer outputs the value of each position in the feature map; is the weight value of the convolution kernel; a, b are position indexes; is the bias of the feature map of the lth layer; σ is the activation function; Multiple convolutional layers are stacked to form a feature pyramid (FPN), where the low-level layers extract texture details (such as crack edges and corrosion patches) and the high-level layers encode global semantics (such as defect types). Based on the extracted features, a candidate region ROI that may contain defects is generated based on the region proposal network; For the ROI that passes the screening, ROI Align is used to accurately extract the features that match the ROI and pass them to the following branches: Classification branch, target category classification prediction, loss function L cls is the cross entropy loss:

[0021] Among them, K is the total number of categories; y k is the one-hot label corresponding to the true category k; P k is the probability distribution of the predicted category k; The positioning branch accurately locates the target, optimizes the IoU (Intersection over Union) of the target bounding box, improves the overlap between the predicted box and the real bounding box, and reduces the positioning loss.L loc for: ; Among them, B true is the true bounding box of the candidate region; B pred is the predicted target bounding box; Mask branch, which generates target instance masks, generates pixel-level masks of specific categories for each ROI (instance segmentation), and calculates mask errors using binary cross entropy loss L mask : ; Among them, M i,j is the true mask at position (i, j); Predict the mask at position (i, j); Output the category (such as crack, corrosion, blockage), location (Bounding Box), confidence score (Confidence Score) and mask (Mask) of each candidate target.

[0022] In this embodiment, the morphological changes of the pipeline are detected based on the RANSAC point cloud data processing algorithm, as follows: Using the pre-processed pipeline point cloud data, extract the reference cross-sectional shape of the current pipeline; The plane segmentation algorithm is applied to divide the point cloud into several two-dimensional sections for frame-by-frame analysis of the pipeline cross-sectional morphology; The random sampling consensus algorithm RANSAC is used to fit the shape and fit the pipe section from the point cloud. According to the fitting result, it is judged whether the section deviates from the ideal shape. The deviation formula is: ; Among them, r ideal is the pipe cross section, r actual is the radius of the fitting section. If d exceeds the set threshold (e.g. d>0.05), it is determined to be a deformation area; Mark the area where the pipeline is deformed, give the deformation type (expansion, contraction, distortion), and output the deformation deviation d.

[0023] In this embodiment, ultrasonic signal defect detection based on Fourier transform and time-frequency analysis is as follows: For the collected ultrasonic or sonar signal x(t), its spectrum characteristics are extracted by fast Fourier transform: ; Where X(f) is the frequency domain signal; is an imaginary unit; f and t represent frequency and time respectively; Analyze the peak position of the echo signal. The echo peak indicates the strong signal generated by the reflection from the material interface. The time interval Δt of different peaks corresponds to the thickness. T : ; where c is the speed of sound in the material; If the time intervals are not uniform, it is considered that there is material delamination or internal cracks; Perform envelope analysis on the echo signal and evaluate the pipe wall strength by the attenuation rate of the signal amplitude with depth:

[0024] Where c is the speed of sound in the material; A0 is the initial signal amplitude; α is the attenuation coefficient; is the propagation distance; if the attenuation A is higher than the threshold, there is a material structure problem; And use short-time Fourier transform (STFT) to decompose the time-frequency characteristics of the signal and locate the specific time point when the pipe wall defect occurs: ; Where w(t) is the window function; is the time-integrated variable; f is the frequency; Locate the defect position based on characteristic frequency and energy changes; Finally, the wall thickness deviation value, delamination or internal crack risk area are output.

[0025] In this embodiment, S4 is specifically: Based on the preprocessed point cloud data, the ICP algorithm is used to align the multi-frame point clouds; The Delaunay triangulation algorithm is used to generate a three-dimensional surface mesh, and each fitted section in the point cloud forms a continuous pipe wall surface through an interpolation algorithm; The preprocessed image is then wrapped onto the generated 3D surface mesh to enhance the texture realism and construct a 3D model of the underground pipeline; Align the underground pipeline 3D model with the geographic coordinate data, obtain the time series of the inspection robot's position coordinates from the IMU and odometer, align them with the pipeline layout reference data in the geographic information system (GIS), and convert the robot's reference coordinate system into the GIS global coordinate system; By using Mask R-CNN to detect the defect location in the image and obtain the bounding box or pixel area, the pixel coordinates of the image detection result are converted into 3D model coordinates; and the point cloud abnormal area is directly matched to the 3D model coordinates; the corresponding 3D coordinates of each ultrasonic sampling point are calculated according to the ultrasonic signal defect detection results; Align all inspection results (3D coordinates and attribute information) to a unified 3D model reference coordinate system of underground pipelines, and mark defect types (such as cracks, corrosion, deformation, and blockage).

[0026] In this embodiment, S5 is specifically: Quantitatively score the defects according to their type and severity to form the pipeline health score PHI; ; in, According to the defect type The severity of the disease is graded and scored; Defect type The weight of; n is the total number of defect types; And evaluate the repair priority based on defect type priority, severity and location distribution; An assessment report is generated based on the test results and model assessment, including basic information, defect distribution summary and comprehensive score PHI; the basic information includes pipeline number, location coordinates, material, length, diameter and service life; the defect distribution summary includes the number, location, type and severity of defects.

[0027] An intelligent detection and evaluation system for underground pipeline defects, comprising a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, the steps in the intelligent detection and evaluation method for underground pipeline defects are specifically performed. It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0028] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0029] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0030] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0031] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

Claims

1. An intelligent detection and evaluation method for underground pipeline defects, characterized in that: The following steps are involved: S1: Based on the inspection robot, collect multimodal data of underground pipelines and transmit them to the intelligent detection and evaluation platform; S2: Intelligent detection and evaluation platform, pre-processing the collected multi-modal data; S3: Based on the pre-processed multimodal data, defect detection of underground pipelines is performed, including: Image recognition algorithms based on convolutional neural networks detect visible defects, including cracks, corrosion, and blockages; Based on the RANSAC point cloud data processing algorithm, the morphological changes of the pipeline are detected and the deformation area is located; Ultrasonic signal defect detection based on Fourier transform and time-frequency analysis; S4: constructing a three-dimensional model of the underground pipeline and superimposing the defect detection results on the three-dimensional model; S5: Based on the 3D model of underground pipelines, intelligently evaluate the underground pipelines and generate a pipeline condition assessment report.

2. The intelligent detection and evaluation method for underground pipeline defects according to claim 1 is characterized in that: The multimodal data includes: image data of the inside of the pipeline obtained by a high-definition camera; three-dimensional point cloud data of the inside of the pipeline constructed by a lidar; depth detection signals for pipe wall thickness detection or potential defect detection obtained by an ultrasonic sensor; humidity and temperature obtained by an environmental sensor; and positioning information of the inspection robot recorded by an IMU and an odometer.

3. The intelligent detection and evaluation method for underground pipeline defects according to claim 2 is characterized in that: The preprocessing is specifically as follows: S21: Remove duplicate, ambiguous or redundant sampling data from multimodal data; S22: Process different types of data separately, including: Image data, use Gaussian filtering to eliminate image noise, enhance pipeline feature details, apply histogram equalization to improve image contrast, and use the camera calibration matrix to eliminate image distortion caused by high-definition cameras; For point cloud data, use the neighborhood filtering method to remove invalid point cloud noise and fill in the missing parts through edge fitting or interpolation; The deep detection signal is de-noised using a Butterworth low-pass filter, and the envelope signal is extracted using the Hilbert transform to capture the waveform anomaly caused by the defect; Humidity and temperature data obtained by environmental sensors, using time series interpolation to fill in missing data and normalize different environmental parameters; S23: Align the multimodal data after being processed separately in time and space to unify the multimodal data format.

4. The intelligent detection and evaluation method for underground pipeline defects according to claim 1 is characterized in that: The convolutional neural network-based image recognition algorithm adopts the Mask R-CNN target detection framework, including a convolutional feature extraction network, a region proposal network, and a ROI alignment and detection network: The convolutional feature extraction network uses convolutional layers to extract features from images. The feature extraction process is as follows: ； in, is the value of the feature map of the lth layer at position (i, j); For the The layer outputs the value of each position in the feature map; is the weight value of the convolution kernel; a, b are position indexes; is the bias of the feature map of the lth layer; σ is the activation function; Multiple convolutional layers are stacked to form a feature pyramid, where the lower layers extract texture details and the higher layers encode global semantics. Based on the extracted features, a candidate region ROI that may contain defects is generated based on the region proposal network; For the ROI that passes the screening, ROI Align is used to accurately extract the features that match the ROI, and the features are passed to the following branches: Classification branch, target category classification prediction, loss function L cls is the cross entropy loss: Among them, K is the total number of categories; y k is the one-hot label corresponding to the true category k; P k is the probability distribution of the predicted category k; The positioning branch accurately locates the target, optimizes the IoU of the target bounding box, improves the overlap between the predicted box and the real bounding box, and reduces the positioning loss. L loc for: ； Among them, B true is the true bounding box of the candidate region; B pred is the predicted target bounding box; The mask branch generates target instance masks, generates a pixel-level mask of a specific category for each ROI, and calculates the mask error using binary cross entropy loss L mask : ； Among them, M i,j is the true mask at position (i, j); Predict the mask at position (i, j); Output the category, location and mask of each candidate object.

5. The intelligent detection and evaluation method for underground pipeline defects according to claim 1 is characterized in that: The RANSAC-based point cloud data processing algorithm detects the morphological changes of the pipeline, as follows: Using the pre-processed pipeline point cloud data, extract the reference cross-sectional shape of the current pipeline; The plane segmentation algorithm is applied to divide the point cloud into several two-dimensional sections for frame-by-frame analysis of the pipeline cross-sectional morphology; The random sampling consensus algorithm RANSAC is used to fit the shape and fit the pipe section from the point cloud. According to the fitting result, it is judged whether the section deviates from the ideal shape. The deviation formula is: ； Among them, r ideal is the pipe cross section, r actual is the radius of the fitting section. If d exceeds the set threshold, it is determined to be a deformation area; Mark the area where the pipeline is deformed, give the deformation type, and output the deformation deviation d.

6. The intelligent detection and evaluation method for underground pipeline defects according to claim 1 is characterized in that: The ultrasonic signal defect detection based on Fourier transform and time-frequency analysis is specifically as follows: For the collected ultrasonic or sonar signal x(t), the signal spectrum characteristics are extracted by fast Fourier transform: ； Where X(f) is the frequency domain signal; is an imaginary unit; f and t represent frequency and time respectively; Analyze the peak position of the echo signal. The echo peak indicates the strong signal generated by the reflection from the material interface. The time interval Δt of different peaks corresponds to the thickness. T : ； where c is the speed of sound in the material; If the time intervals are not uniform, it is considered that there is material delamination or internal cracks; Perform envelope analysis on the echo signal and evaluate the pipe wall strength by the attenuation rate of the signal amplitude with depth: Where A0 is the initial signal amplitude; α is the attenuation coefficient; is the propagation distance; if the attenuation A is higher than the threshold, there is a material structure problem; And use short-time Fourier transform (STFT) to decompose the time-frequency characteristics of the signal and locate the specific time point when the pipe wall defect occurs: ; Where w(t) is the window function; is the time-integrated variable; f is the frequency; Locate the defect position based on characteristic frequency and energy changes; Finally, the wall thickness deviation value, delamination or internal crack risk area are output.

7. The intelligent detection and evaluation method for underground pipeline defects according to claim 1 is characterized in that: The S4 is specifically: Based on the preprocessed point cloud data, the ICP algorithm is used to align the multi-frame point clouds; The Delaunay triangulation algorithm is used to generate a three-dimensional surface mesh, and each fitted section in the point cloud forms a continuous pipe wall surface through an interpolation algorithm; The preprocessed image is then wrapped onto the generated 3D surface mesh to enhance the texture realism and construct a 3D model of the underground pipeline; Align the underground pipeline 3D model with the geographic coordinate data, obtain the time series of the inspection robot's position coordinates from the IMU and odometer, align them according to the pipeline layout reference data in the geographic information system GIS, and convert the robot's reference coordinate system into the GIS global coordinate system; By using Mask R-CNN to detect the defect location in the image and obtain the bounding box or pixel area, the pixel coordinates of the image detection result are converted into 3D model coordinates; and the point cloud abnormal area is directly matched to the 3D model coordinates; the corresponding 3D coordinates of each ultrasonic sampling point are calculated according to the ultrasonic signal defect detection results; All inspection results are aligned to a unified underground pipeline 3D model reference coordinate system and the defect types are marked.

8. The intelligent detection and evaluation method for underground pipeline defects according to claim 1 is characterized in that: The S5 is specifically: Quantitatively score the defects according to their type and severity to form the pipeline health score PHI; ； in, According to the defect type The severity of the disease is graded and scored; Defect type The weight of; n is the total number of defect types; And evaluate the repair priority based on defect type priority, severity and location distribution; An assessment report is generated based on the test results and model assessment, including basic information, defect distribution summary and comprehensive score PHI; the basic information includes pipeline number, location coordinates, material, length, diameter and service life; the defect distribution summary includes the number, location, type and severity of defects.

9. An intelligent detection and evaluation system for underground pipeline defects, characterized in that: The method comprises a processor, a memory and a computer program stored in the memory. When the processor executes the computer program, the method specifically executes the steps in the intelligent detection and evaluation method for underground pipeline defects as described in any one of claims 1 to 8.

Images