Urban pipeline defect detection method and system based on AI identification

Abstract

The invention provides an urban pipeline defect detection method and system based on AI identification, and the method comprises the steps: carrying out the frame extraction of a collected video of a received pipeline inner diameter video collection operation of a pipeline robot, obtaining a sample image, numbering the sample image, and inserting the numbered sample image into a to-be-detected task queue; extracting a sample image corresponding to the first serial number in the to-be-detected task queue, and submitting the images to the detection reasoning model one by one for reasoning detection; receiving an identification result returned by the detection reasoning model; and performing defect position drawing and identification on the sample image based on an identification result. According to the invention, through analysis of urban pipeline detection CCTV video data and marking classification of defect images, a defect identification AI model is formed by training pipeline defect target features by means of a neural convolutional network algorithm, and automatic identification of pipeline quality management is realized through intelligent analysis of pipeline quality, so that manual intervention can be reduced, working efficiency can be improved, and labor intensity can be reduced. And finally, a unified standard library for pipeline quality detection can be constructed and formed.

Description

Technical Field

[0001] The present invention relates to the technical field of garbage cleaning and recycling, and in particular to an urban pipeline defect detection method and system based on AI recognition. Background Art

[0002] As an important infrastructure of the city, the drainage network plays a very important role in the city's environmental protection and production process. During the construction of the existing pipelines, there will be construction defects such as pipe misalignment and rupture. In addition, as the use time of the existing pipelines in a large number of cities increases, the problems of pipeline deformation, water seepage, and damage become more and more serious. In order to improve the drainage capacity of the pipeline, it is necessary to organize professional forces to conduct inspections and regular maintenance in a timely manner. At present, in the field of drainage pipeline quality inspection and defect detection, there are mainly the following existing technologies:

[0003] 1. Traditional manual inspection method: Traditional inspection of drainage pipe networks mainly relies on a large amount of manpower. Inspectors need to enter the pipe or use specific equipment to observe and measure outside the pipe. This method has a complicated operation process and consumes a lot of time and manpower. In addition, the inspection results are easily affected by human subjective experience factors. Different inspectors may come to different judgments on the same pipe condition. It is difficult to unify management and quality standards, and it is difficult to meet the overall development requirements of the project.

[0004] 2. Some automated testing equipment: Although there are some automated testing equipment, these devices are usually single-function and can only detect specific types of defects, such as simple pipeline deformation detection equipment. In addition, these devices have limited data processing capabilities and cannot conduct comprehensive and in-depth analysis and classification of the detected data. Summary of the invention

[0005] One of the purposes of the present invention is to provide a method and system for urban pipeline defect detection based on AI recognition. By analyzing CCTV video data of urban pipeline detection and labeling and classifying defect images, a defect recognition AI model is formed by training the target features of pipeline defects with the help of a neural convolutional network algorithm (CNN). Automatic recognition of pipeline quality management is achieved through intelligent analysis of pipeline quality, which can reduce manual intervention and improve work efficiency. Finally, a unified standard library for pipeline quality detection can be established.

[0006] An embodiment of the present invention provides a method for detecting urban pipeline defects based on AI recognition, comprising:

[0007] Extract frames from the received video of the pipeline inner diameter video acquisition operation of the pipeline robot to obtain sample images, and number the sample images and insert them into the task queue to be detected;

[0008] Extract the sample image corresponding to the first number in the task queue to be detected and submit the images one by one to the detection inference model for inference detection;

[0009] Receive the recognition result returned by the detection inference model;

[0010] Based on the recognition results, the defect locations of the sample images are drawn and marked.

[0011] Preferably, the pipeline robot comprises: a walking mechanism, a device body, an image acquisition device, a lighting device, a communication module and a control module;

[0012] Among them, the control module is electrically connected to the walking mechanism, the image acquisition device, the lighting device, and the communication module respectively.

[0013] Preferably, the walking mechanism comprises: a plurality of telescopic connectors and walking wheels arranged at the ends of the telescopic connectors;

[0014] The device body is configured in a ring shape; a plurality of telescopic connectors are evenly distributed on the periphery of the device body; an illumination device and an image acquisition device are arranged at the rotating end of a dual-axis gimbal on a counterweight platform of the inner ring of the device body; the rotation of the dual-axis gimbal changes the shooting direction of the image acquisition device and the illumination device; a control module and a communication module are arranged in the counterweight platform;

[0015] A guide rail is arranged on the periphery of the equipment body and at the connection position between the equipment body and the telescopic connecting body. One end of the telescopic connecting body is hingedly arranged on a platform sliding on the guide rail through a hinge structure; under the action of the hinge structure, the telescopic connecting body can be opened to the plane corresponding to the ring of the equipment body, and can also be folded to the vertical direction of the plane corresponding to the ring of the equipment body.

[0016] Preferably, the urban pipeline defect detection method based on AI recognition also includes:

[0017] When the pipeline robot performs the video acquisition operation of the pipeline inner diameter, the acquired video is analyzed to obtain the control analysis result;

[0018] Controlling the motion of the pipeline robot based on the control analysis results;

[0019] The collected video is analyzed to obtain control analysis results, including:

[0020] Extract the image of the current frame, construct a coordinate system with the center position of the image as the origin, and analyze the image to determine the coordinates of the central axis point of the pipeline;

[0021] Calculate the distance between the coordinates of the central axis point and the coordinate origin;

[0022] When the distance is greater than a preset trigger threshold, the image acquisition device enters a position correction mode;

[0023] Constructing a modified analysis data set based on the distances of a preset number of consecutive frames of images;

[0024] Based on the correction analysis data set, the correction parameters of the dual-axis gimbal are determined from the pre-configured correction library.

[0025] Preferably, analyzing the collected video to obtain control analysis results also includes:

[0026] Obtain a three-dimensional pipeline map and locate the pipeline robot in the three-dimensional pipeline map;

[0027] Match the image of the current frame with the trigger image in the pre-configured speed adjustment trigger library;

[0028] When the trigger image matched by the image of the current frame is different from the trigger image matched by the image of the previous frame, the speed adjustment mode is entered;

[0029] Arranging images of a first number of consecutive frames preset after entering the speed adjustment mode in order to form an analysis image set;

[0030] Retrieving each standard image set in a pre-configured speed adjustment library, and sequentially using the retrieved standard image sets to match each image in the standard image set with each image in the analysis image set;

[0031] When the matching result is consistent, the speed adjustment parameters corresponding to the standard image set are retrieved.

[0032] Preferably, analyzing the collected video to obtain control analysis results also includes:

[0033] Analyze the image of the current frame to determine the sewage boundary;

[0034] Based on the sewage boundary, the guide rail is controlled to drive the platform to move;

[0035] Among them, based on the sewage boundary, the guide rail movement is controlled to drive the platform to move, including:

[0036] Construct pipe cross-sections and annotate wastewater boundaries, and determine the length of wastewater boundaries;

[0037] Based on the length of the sewage boundary, a pre-configured control table is queried to determine the angle between the telescopic connectors located on both sides of the counterweight platform;

[0038] Control the rail movement according to the included angle and the pre-configured rail control table.

[0039] The present invention also provides an urban pipeline defect detection system based on AI recognition, comprising: a pipeline robot, a control terminal and an analysis server;

[0040] The analysis server performs the following operations:

[0041] Extract frames from the received video of the pipeline inner diameter video acquisition operation of the pipeline robot to obtain sample images, and number the sample images and insert them into the task queue to be detected;

[0042] Extract the sample image corresponding to the first number in the task queue to be detected and submit the images one by one to the detection inference model for inference detection;

[0043] Receive the recognition result returned by the detection inference model;

[0044] Based on the recognition results, the defect locations of the sample images are drawn and marked.

[0045] Preferably, the pipeline robot comprises: a walking mechanism, a device body, an image acquisition device, a lighting device, a communication module and a control module;

[0046] Among them, the control module is electrically connected to the walking mechanism, the image acquisition device, the lighting device, and the communication module respectively.

[0047] Preferably, the walking mechanism comprises: a plurality of telescopic connectors and walking wheels arranged at the ends of the telescopic connectors;

[0048] The device body is configured in a ring shape; a plurality of telescopic connectors are evenly distributed on the periphery of the device body; an illumination device and an image acquisition device are arranged at the rotating end of a dual-axis gimbal on a counterweight platform of the inner ring of the device body; the rotation of the dual-axis gimbal changes the shooting direction of the image acquisition device and the illumination device; a control module and a communication module are arranged in the counterweight platform;

[0049] A guide rail is arranged on the periphery of the equipment body and at the connection position between the equipment body and the telescopic connecting body. One end of the telescopic connecting body is hingedly arranged on a platform sliding on the guide rail through a hinge structure; under the action of the hinge structure, the telescopic connecting body can be opened to the plane corresponding to the ring of the equipment body, and can also be folded to the vertical direction of the plane corresponding to the ring of the equipment body.

[0050] Preferably, the control module performs the following operations:

[0051] When the pipeline robot performs the video acquisition operation of the pipeline inner diameter, the acquired video is analyzed to obtain the control analysis result;

[0052] Controlling the motion of the pipeline robot based on the control analysis results;

[0053] The collected video is analyzed to obtain control analysis results, including:

[0054] Extract the image of the current frame, construct a coordinate system with the center position of the image as the origin, and analyze the image to determine the coordinates of the central axis point of the pipeline;

[0055] Calculate the distance between the coordinates of the central axis point and the coordinate origin;

[0056] When the distance is greater than a preset trigger threshold, the image acquisition device enters a position correction mode;

[0057] Constructing a modified analysis data set based on the distances of a preset number of consecutive frames of images;

[0058] Based on the correction analysis data set, the correction parameters of the dual-axis gimbal are determined from the pre-configured correction library.

[0059] Preferably, analyzing the collected video to obtain control analysis results also includes:

[0060] Obtain a three-dimensional pipeline map and locate the pipeline robot in the three-dimensional pipeline map;

[0061] Match the image of the current frame with the trigger image in the pre-configured speed adjustment trigger library;

[0062] When the trigger image matched by the image of the current frame is different from the trigger image matched by the image of the previous frame, the speed adjustment mode is entered;

[0063] Arranging images of a first number of consecutive frames preset after entering the speed adjustment mode in order to form an analysis image set;

[0064] Retrieving each standard image set in a pre-configured speed adjustment library, and sequentially using the retrieved standard image sets to match each image in the standard image set with each image in the analysis image set;

[0065] When the matching result is consistent, the speed adjustment parameters corresponding to the standard image set are retrieved.

[0066] Preferably, analyzing the collected video to obtain control analysis results also includes:

[0067] Analyze the image of the current frame to determine the sewage boundary;

[0068] Based on the sewage boundary, the guide rail is controlled to drive the platform to move;

[0069] Among them, based on the sewage boundary, the guide rail movement is controlled to drive the platform to move, including:

[0070] Construct pipe cross-sections and annotate wastewater boundaries, and determine the length of wastewater boundaries;

[0071] Based on the length of the sewage boundary, a pre-configured control table is queried to determine the angle between the telescopic connectors located on both sides of the counterweight platform;

[0072] Control the rail movement according to the included angle and the pre-configured rail control table.

[0073] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0074] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0076] Figure 1 Schematic diagram of a method for detecting urban pipeline defects based on AI recognition in an embodiment of the present invention;

[0077] Figure 2 A schematic diagram of a pipeline robot according to an embodiment of the present invention;

[0078] Figure 3 A partial cross-sectional view of the equipment body of the pipeline robot in an embodiment of the present invention;

[0079] Figure 4 Schematic diagram of a system for detecting urban pipeline defects based on AI recognition in an embodiment of the present invention;

[0080] Figure 5 This is a schematic diagram of data transmission of an urban pipeline defect detection system based on AI recognition in an embodiment of the present invention. DETAILED DESCRIPTION

[0081] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0082] The embodiment of the present invention provides a method for detecting urban pipeline defects based on AI recognition, such as Figure 1 As shown, including:

[0083] Step 1: extract frames from the collected video of the pipeline inner diameter video collection operation of the pipeline robot to obtain sample images, and number the sample images and insert them into the task queue to be detected;

[0084] Step 2: extract the sample image corresponding to the first number in the task queue to be detected and submit the images one by one to the detection inference model for inference detection;

[0085] Step 3: Receive the recognition result returned by the detection inference model;

[0086] Step 4: Based on the recognition results, the defect positions of the sample images are drawn and marked.

[0087] The specific scenario is as follows: the operator at the construction site uses the terminal PAD (control terminal) to operate the pipeline robot to collect video of the inner diameter of the pipeline; after the pipeline robot completes the collection, it automatically generates a video file and transmits it to the video analysis service; the video analysis service (analysis server) saves the video file, creates a task information file based on the video file MD5 as the file name, and saves the initial state value as the frame to be extracted; the video file is extracted at 3 frames per second, and the extracted frame images are stored in the detection sample folder with the video MD5 as the file name. Each frame of the image is stored in the current video detection sample folder as a video detection sample data set in the format of {video file MD5_video image frame capture time in milliseconds.png}. After the video frame extraction is completed, the task information file status value is modified to be detected; and the MD5 of the video to be detected is inserted into the task queue to be detected as the task ID; the video analysis service obtains the task information ID at the top of the current queue in the queue to be detected, searches for the video detection sample data set with MD5 as the folder name according to the MD5 data of the task information ID, and submits the images one by one to the detection reasoning model service for reasoning detection according to the file name of the video frame extracted image, and waits for the reasoning model service to return the recognition result; according to the recognition result, if there is a detection target recognition item, record the detection image frame file name, target defect code, detection target matrix upper left corner in the image coordinates (p1), detection target image width (w), detection target image height (h); according to the video analysis recognition result, the detection image frame is drawn and marked according to the coordinates (p1), image width (w), and image height (h); the detection reasoning model is trained and converged in advance, and is used to monitor pipeline defects through internal pipeline images.

[0088] Among them, the pipeline robot includes: a walking mechanism, a device body, an image acquisition device, a lighting device, a communication module and a control module;

[0089] Among them, the control module is electrically connected to the walking mechanism, the image acquisition device, the lighting device, and the communication module respectively.

[0090] The pipeline robot moves in the pipeline through the walking mechanism, and the image acquisition device acquires images in the pipeline. When the image acquisition device acquires video images in the pipeline, the lighting device illuminates, and the acquired images are sent to the staff's control terminal through the communication module, and then sent to the server for processing.

[0091] In one embodiment, Figure 2 and Figure 3 As shown, the walking mechanism 11 includes: a plurality of telescopic connecting bodies 111 and walking wheels 112 arranged at the ends of the telescopic connecting bodies 111;

[0092] The device body 12 is configured in a ring shape; a plurality of telescopic connectors 111 are evenly distributed on the outer periphery of the device body 12; the lighting device 14 and the image acquisition device 13 are arranged at the rotating end of the dual-axis gimbal 16 on the counterweight platform 15 of the inner ring of the device body 12; the rotation of the dual-axis gimbal 16 changes the shooting direction of the image acquisition device 13 and the lighting device 14; the control module and the communication module are arranged in the counterweight platform 15;

[0093] A guide rail 113 is provided on the periphery of the device body 12 and at the connection position between the device body 12 and the telescopic connecting body 111. One end of the telescopic connecting body 111 is hingedly provided on a platform 114 that slides on the guide rail 113 through a hinge structure. The telescopic connecting body 111 can be opened to the plane corresponding to the ring of the device body 12 under the action of the hinge structure, and can also be folded to the vertical direction of the plane corresponding to the ring of the device body 12.

[0094] The telescopic connector 111 is opened to the plane corresponding to the ring of the device body 12 through the hinge structure for convenient use, so as to realize the movement of the pipeline robot in the pipeline; when folded to the vertical direction of the plane corresponding to the ring of the device body 12, it can be placed on the ground with the walking wheels 112 when not in use. After being placed on the ground, the image acquisition device 13 and the lighting device 14 are far away from the ground, which can avoid bumping these components; the dual-axis pan-tilt 16 realizes the adjustment of the acquisition direction of the image acquisition device 13; after moving through the pipeline, the image acquisition direction of the image acquisition device 13 can be directly turned 180 degrees by the dual-axis pan-tilt 16 without the robot turning around, and the image video can be collected again on the return journey; the pipeline can be inspected from two directions. It is not easy to find problems in some locations by shooting videos from one direction; the accuracy and comprehensiveness of the detection are guaranteed by shooting video images twice.

[0095] In order to enable the pipeline robot to perform adaptive work adjustments according to actual conditions during the video sampling process, in one embodiment, the urban pipeline defect detection method based on AI recognition also includes:

[0096] When the pipeline robot performs the video acquisition operation of the pipeline inner diameter, the acquired video is analyzed to obtain the control analysis result;

[0097] Controlling the motion of the pipeline robot based on the control analysis results;

[0098] The collected video is analyzed to obtain control analysis results, including:

[0099] Extract the image of the current frame, construct a coordinate system with the center position of the image as the origin, and analyze the image to determine the coordinates of the central axis point of the pipeline;

[0100] Calculate the distance between the coordinates of the central axis point and the coordinate origin;

[0101] When the distance is greater than a preset trigger threshold, the image acquisition device enters a position correction mode;

[0102] Constructing a modified analysis data set based on the distances of a preset number of consecutive frames of images;

[0103] Based on the correction analysis data set, the correction parameters of the dual-axis gimbal are determined from the pre-configured correction library.

[0104] This implementation analyzes the coincidence between the center position of the image and the central axis point of the pipeline, and makes adjustments to ensure the distribution of the pipeline in the image, which is convenient for subsequent image analysis. When making corrections, the distance between the coordinates of the central axis point (i.e., the point where the center of the cross section of the pipeline is located) and the coordinate origin is calculated and compared with the trigger threshold. The trigger threshold and the situation below it indicate that the image offset is not large, which may be caused by the unbalanced walking of the walking mechanism, and no correction is required. However, when the trigger threshold is exceeded, adjustment is required. When making adjustments, the correction library configured in advance is used. In the correction library, the correction parameters are associated with the correction analysis data set in a one-to-one correspondence. The correction data set is to arrange the distances calculated for a preset number of frames of images (for example, any value from 10 to 30) in sequence. The continuous frame images are used to analyze the changing relationship between the continuous distances, so as to determine the applicability of the adjusted correction parameters and exclude the influence of other factors (the position offset of the image acquisition device when the walking mechanism happens to walk to a bulge or crack in the pipeline) on the correction.

[0105] In one embodiment, analyzing the collected video to obtain control analysis results also includes:

[0106] Obtain a three-dimensional pipeline map and locate the pipeline robot in the three-dimensional pipeline map;

[0107] Match the image of the current frame with the trigger image in the pre-configured speed adjustment trigger library; the speed adjustment trigger library contains different trigger images, i.e., trigger images corresponding to different moving speeds are pre-configured, and each trigger image corresponds to a different environment; for example, the pipeline is divided into different environments according to the type and degree of dirt on the pipeline; the types include: sludge, oil stains, moss, etc.;

[0108] When the trigger image matched by the current frame image is different from the trigger image matched by the previous frame image, the speed adjustment mode is entered; when the trigger image matched by the current frame image is different from the trigger image matched by the previous frame image, it indicates that the environment has changed;

[0109] Arranging a first number of consecutive preset frames (any number from 5 to 30) of images after entering the speed adjustment mode in order to form an analysis image set;

[0110] Retrieving each standard image set in a pre-configured speed adjustment library, and sequentially using the retrieved standard image sets to match each image in the standard image set with each image in the analysis image set;

[0111] When the matching result is consistent, the speed adjustment parameter associated with the standard image set is retrieved, that is, the similarity between each image in the standard image set and each image in the analysis image set is calculated; when the similarities are greater than or equal to the preset threshold, the matching is determined to be consistent.

[0112] In addition, there is often sewage or other water in the pipeline. In order to ensure that the pipeline robot can continue to work, in one embodiment, the collected video is analyzed to obtain the control analysis results, which also includes:

[0113] Analyze the image of the current frame to determine the sewage boundary;

[0114] Based on the sewage boundary, the guide rail is controlled to drive the platform to move;

[0115] Among them, based on the sewage boundary, the guide rail movement is controlled to drive the platform to move, including:

[0116] Construct pipe cross-sections and annotate wastewater boundaries, and determine the length of wastewater boundaries;

[0117] Based on the length of the sewage boundary, a pre-configured control table is queried to determine the angle between the telescopic connectors located on both sides of the counterweight platform;

[0118] Control the rail movement according to the included angle and the pre-configured rail control table.

[0119] In addition, in order to achieve operational stability, an angle threshold can be configured. When the angle determined by the sewage boundary is greater than or equal to the angle threshold, the pipeline robot is controlled to transfer out of the pipeline.

[0120] This embodiment determines the sewage situation by analyzing the image, and adjusts the angle between the two lower parts of the walking mechanism to achieve running on the sewage;

[0121] Since a large angle between the two lower parts of the walking mechanism will have an adverse effect on the operation of the robot, it is necessary to comprehensively analyze the historical detection data of the water level in the pipe network when working; analyze the historical detection data and determine the pipes with water levels less than or equal to the preset threshold as detection objects.

[0122] The present invention also provides an urban pipeline defect detection system based on AI recognition, such as Figure 4 As shown, it includes: a pipeline robot 1, a control terminal 2 and an analysis server 3;

[0123] The analysis server 3 performs the following operations:

[0124] Extract frames from the received video of the pipeline inner diameter video acquisition operation of the pipeline robot to obtain sample images, and number the sample images and insert them into the task queue to be detected;

[0125] Extract the sample image corresponding to the first number in the task queue to be detected and submit the images one by one to the detection inference model for inference detection;

[0126] Receive the recognition result returned by the detection inference model;

[0127] Based on the recognition results, the defect locations of the sample images are drawn and marked. Figure 5The data transmission diagram corresponding to the system; the operator at the construction site uses the terminal PAD (control terminal) to operate the pipeline robot to collect the video of the inner diameter of the pipeline; after the pipeline robot completes the collection, it automatically generates a video file and transmits it to the video analysis service; the video analysis service (analysis server) saves the video file, creates a task information file based on the video file MD5 as the file name, and saves the initial state value as the frame to be extracted; the video file is extracted at 3 frames per second, and the extracted frame pictures are stored in the detection sample folder with the video MD5 as the file name. Each frame of the picture is stored in the current video detection sample folder as a video detection sample data set in the format of {video file MD5_video picture frame capture time in milliseconds.png}. After the video frame extraction is completed, the task information file status value is modified to be detected; and the MD5 of the video to be detected is inserted into the task queue to be detected as the task ID; the video analysis service obtains the task information ID of the first task in the current queue to be detected, and searches for the video detection sample data set with MD5 as the folder name according to the MD5 data of the task information ID, and submits the pictures one by one to the detection reasoning model service for reasoning detection according to the file name of the video frame extraction picture, and waits for the reasoning model service to return the recognition result; according to the recognition result, if there is a detection target recognition item, record the detection picture frame file name, target defect code, the upper left corner of the detection target matrix in the picture coordinates (p1), the detection target picture width (w), and the detection target picture height (h); according to the video analysis recognition result, the defect position of the detection picture frame is drawn and marked according to the coordinates (p1), picture width (w), and picture height (h).

[0128] Among them, the pipeline robot includes: a walking mechanism, a device body, an image acquisition device, a lighting device, a communication module and a control module;

[0129] Among them, the control module is electrically connected to the walking mechanism, the image acquisition device, the lighting device, and the communication module respectively.

[0130] Wherein, the walking mechanism comprises: a plurality of telescopic connectors and walking wheels arranged at the ends of the telescopic connectors;

[0131] The device body is configured in a ring shape; a plurality of telescopic connectors are evenly distributed on the periphery of the device body; an illumination device and an image acquisition device are arranged at the rotating end of a dual-axis gimbal on a counterweight platform of the inner ring of the device body; the rotation of the dual-axis gimbal changes the shooting direction of the image acquisition device and the illumination device; a control module and a communication module are arranged in the counterweight platform;

[0132] A guide rail is arranged on the periphery of the equipment body and at the connection position between the equipment body and the telescopic connecting body. One end of the telescopic connecting body is hingedly arranged on a platform sliding on the guide rail through a hinge structure; under the action of the hinge structure, the telescopic connecting body can be opened to the plane corresponding to the ring of the equipment body, and can also be folded to the vertical direction of the plane corresponding to the ring of the equipment body.

[0133] The control module performs the following operations:

[0134] When the pipeline robot performs the video acquisition operation of the pipeline inner diameter, the acquired video is analyzed to obtain the control analysis result;

[0135] Controlling the motion of the pipeline robot based on the control analysis results;

[0136] The collected video is analyzed to obtain control analysis results, including:

[0137] Extract the image of the current frame, construct a coordinate system with the center position of the image as the origin, and analyze the image to determine the coordinates of the central axis point of the pipeline;

[0138] Calculate the distance between the coordinates of the central axis point and the coordinate origin;

[0139] When the distance is greater than a preset trigger threshold, the image acquisition device enters a position correction mode;

[0140] Constructing a modified analysis data set based on the distances of a preset number of consecutive frames of images;

[0141] Based on the correction analysis data set, the correction parameters of the dual-axis gimbal are determined from the pre-configured correction library.

[0142] The collected video is analyzed to obtain the control analysis results, which also includes:

[0143] Obtain a three-dimensional pipeline map and locate the pipeline robot in the three-dimensional pipeline map;

[0144] Match the image of the current frame with the trigger image in the pre-configured speed adjustment trigger library;

[0145] When the trigger image matched by the image of the current frame is different from the trigger image matched by the image of the previous frame, the speed adjustment mode is entered;

[0146] Arranging images of a first number of consecutive frames preset after entering the speed adjustment mode in order to form an analysis image set;

[0147] Retrieving each standard image set in a pre-configured speed adjustment library, and sequentially using the retrieved standard image sets to match each image in the standard image set with each image in the analysis image set;

[0148] When the matching result is consistent, the speed adjustment parameters corresponding to the standard image set are retrieved.

[0149] The collected video is analyzed to obtain the control analysis results, which also includes:

[0150] Analyze the image of the current frame to determine the sewage boundary;

[0151] Based on the sewage boundary, the guide rail is controlled to drive the platform to move;

[0152] Among them, based on the sewage boundary, the guide rail movement is controlled to drive the platform to move, including:

[0153] Construct pipe cross-sections and annotate wastewater boundaries, and determine the length of wastewater boundaries;

[0154] Based on the length of the sewage boundary, a pre-configured control table is queried to determine the angle between the telescopic connectors located on both sides of the counterweight platform;

[0155] Control the rail movement according to the included angle and the pre-configured rail control table.

[0156] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for detecting urban pipeline defects based on AI recognition, characterized in that: include: Extract frames from the received video of the pipeline inner diameter video acquisition operation of the pipeline robot to obtain sample images, and number the sample images and insert them into the task queue to be detected; Extract the sample image corresponding to the first number in the task queue to be detected and submit the images one by one to the detection inference model for inference detection; Receive the recognition result returned by the detection inference model; Based on the recognition results, the defect locations of the sample images are drawn and marked.

2. The urban pipeline defect detection method based on AI recognition according to claim 1 is characterized in that: The pipeline robot includes: a walking mechanism, a device body, an image acquisition device, a lighting device, a communication module and a control module; Among them, the control module is electrically connected to the walking mechanism, the image acquisition device, the lighting device, and the communication module respectively.

3. The urban pipeline defect detection method based on AI recognition according to claim 2 is characterized in that: The walking mechanism comprises: a plurality of telescopic connectors and walking wheels arranged at the ends of the telescopic connectors; The device body is configured in a ring shape; a plurality of telescopic connectors are evenly distributed on the periphery of the device body; an illumination device and an image acquisition device are arranged at the rotating end of a dual-axis gimbal on a counterweight platform of the inner ring of the device body; the rotation of the dual-axis gimbal changes the shooting direction of the image acquisition device and the illumination device; a control module and a communication module are arranged in the counterweight platform; A guide rail is arranged on the periphery of the equipment body and at the connection position between the equipment body and the telescopic connecting body. One end of the telescopic connecting body is hingedly arranged on a platform sliding on the guide rail through a hinge structure; under the action of the hinge structure, the telescopic connecting body can be opened to the plane corresponding to the ring of the equipment body, and can also be folded to the vertical direction of the plane corresponding to the ring of the equipment body.

4. The urban pipeline defect detection method based on AI recognition as claimed in claim 3 is characterized in that: Also includes: When the pipeline robot performs the video acquisition operation of the pipeline inner diameter, the acquired video is analyzed to obtain the control analysis result; Controlling the motion of the pipeline robot based on the control analysis results; The collected video is analyzed to obtain control analysis results, including: Extract the image of the current frame, construct a coordinate system with the center position of the image as the origin, and analyze the image to determine the coordinates of the central axis point of the pipeline; Calculate the distance between the coordinates of the central axis point and the coordinate origin; When the distance is greater than a preset trigger threshold, the image acquisition device enters a position correction mode; Constructing a modified analysis data set based on the distances of a preset number of consecutive frames of images; Based on the correction analysis data set, the correction parameters of the dual-axis gimbal are determined from the pre-configured correction library.

5. The urban pipeline defect detection method based on AI recognition as claimed in claim 4 is characterized in that: Analyze the collected video and obtain the control analysis results, which also includes: Obtain a three-dimensional pipeline map and locate the pipeline robot in the three-dimensional pipeline map; Match the image of the current frame with the trigger image in the pre-configured speed adjustment trigger library; When the trigger image matched by the image of the current frame is different from the trigger image matched by the image of the previous frame, the speed adjustment mode is entered; Arranging images of a first number of consecutive frames preset after entering the speed adjustment mode in order to form an analysis image set; Retrieving each standard image set in a pre-configured speed adjustment library, and sequentially using the retrieved standard image sets to match each image in the standard image set with each image in the analysis image set; When the matching result is consistent, the speed adjustment parameters corresponding to the standard image set are retrieved.

6. The urban pipeline defect detection method based on AI recognition according to claim 4 is characterized in that: Analyze the collected video and obtain the control analysis results, which also includes: Analyze the image of the current frame to determine the sewage boundary; Based on the sewage boundary, the guide rail is controlled to drive the platform to move; Among them, based on the sewage boundary, the guide rail movement is controlled to drive the platform to move, including: Construct pipe cross-sections and annotate wastewater boundaries, and determine the length of wastewater boundaries; Based on the length of the sewage boundary, a pre-configured control table is queried to determine the angle between the telescopic connectors located on both sides of the counterweight platform; Control the rail movement according to the included angle and the pre-configured rail control table.

7. An urban pipeline defect detection system based on AI recognition, characterized in that: include: Pipeline robots, control terminals, and analysis servers; The analysis server performs the following operations: Extract frames from the received video of the pipeline inner diameter video acquisition operation of the pipeline robot to obtain sample images, and number the sample images and insert them into the task queue to be detected; Extract the sample image corresponding to the first number in the task queue to be detected and submit the images one by one to the detection inference model for inference detection; Receive the recognition result returned by the detection inference model; Based on the recognition results, the defect locations of the sample images are drawn and marked.

8. The urban pipeline defect detection system based on AI recognition according to claim 7 is characterized in that: The pipeline robot includes: a walking mechanism, a device body, an image acquisition device, a lighting device, a communication module and a control module; Among them, the control module is electrically connected to the walking mechanism, the image acquisition device, the lighting device, and the communication module respectively.

9. The urban pipeline defect detection system based on AI recognition according to claim 8, characterized in that: The walking mechanism comprises: a plurality of telescopic connectors and walking wheels arranged at the ends of the telescopic connectors; The device body is configured in a ring shape; a plurality of telescopic connectors are evenly distributed on the periphery of the device body; an illumination device and an image acquisition device are arranged at the rotating end of a dual-axis gimbal on a counterweight platform of the inner ring of the device body; the rotation of the dual-axis gimbal changes the shooting direction of the image acquisition device and the illumination device; a control module and a communication module are arranged in the counterweight platform; A guide rail is arranged on the periphery of the equipment body and at the connection position between the equipment body and the telescopic connecting body. One end of the telescopic connecting body is hingedly arranged on a platform sliding on the guide rail through a hinge structure; under the action of the hinge structure, the telescopic connecting body can be opened to the plane corresponding to the ring of the equipment body, and can also be folded to the vertical direction of the plane corresponding to the ring of the equipment body.

10. The urban pipeline defect detection system based on AI recognition according to claim 9, characterized in that: The control module performs the following operations: When the pipeline robot performs the video acquisition operation of the pipeline inner diameter, the acquired video is analyzed to obtain the control analysis result; Controlling the motion of the pipeline robot based on the control analysis results; The collected video is analyzed to obtain control analysis results, including: Extract the image of the current frame, construct a coordinate system with the center position of the image as the origin, and analyze the image to determine the coordinates of the central axis point of the pipeline; Calculate the distance between the coordinates of the central axis point and the coordinate origin; When the distance is greater than a preset trigger threshold, the image acquisition device enters a position correction mode; Constructing a modified analysis data set based on the distances of a preset number of consecutive frames of images; Based on the correction analysis data set, the correction parameters of the dual-axis gimbal are determined from the pre-configured correction library.

Images