Pipeline CCTV detection device and method under complex working conditions

Abstract

The invention discloses a pipeline CCTV detection device and method under a complex working condition, and the device comprises a ship body platform for bearing a detection instrument, and is integrated with an active propulsion system, a hole traction system, a variable buoyancy system, an obstacle clearing mechanism, a multi-sensor system and an AI image recognition module. After intercepting, pumping and ventilating the pipeline, arranging a traction system, putting a ship body platform carrying a detection instrument into a well bottom, and enabling the ship body to advance in the pipeline through cooperative traction and self-propelling power of an intelligent power controller; multiple sensors are used for collecting environment data, video streams are analyzed in real time through the AI image recognition module, and the structural and functional defects of the pipeline are automatically recognized; when blockage is recognized or the estimated resistance is too large, a progressive obstacle response strategy is triggered to overcome obstacles; the problems that traditional detection is insufficient in power, prone to clamping stagnation and low in intelligent level under complex working conditions are solved, and efficient, continuous and automatic pipeline detection under the non-desilting condition is achieved.

Description

Technical Field

[0001] This invention relates to the field of municipal pipeline inspection technology, specifically designing a pipeline CCTV inspection device and method under complex working conditions, which is mainly applied to CCTV inspection of sewage pipelines, culverts, drainage pipelines and similar projects in the field of municipal engineering. Background Technology

[0002] When sewage pipes and culverts in municipal engineering projects experience problems such as poor drainage or blockages, they must be inspected to understand their structural condition (e.g., pipe joints, pipe walls, and foundations) and functional condition (e.g., deposits and scale buildup). Currently, closed-circuit television (CCTV) systems are the mainstream inspection tool for pipelines; however, conventional CCTV inspection methods and devices reveal a series of technical shortcomings that urgently need to be addressed when applied to complex conditions, such as when pipelines have not been thoroughly cleaned, or when there is water accumulation or obstructions inside. Insufficient propulsion and crude control: In pipelines with silt or obstacles, the wheels of conventional CCTV inspection vehicles are prone to slipping and sinking, making it impossible to move. Current methods, which rely entirely on manual towing, depend entirely on the experience and physical strength of the operators, resulting in fatal flaws such as unstable traction, inaccurate speed control, and the inability to quantify resistance. When encountering significant resistance, forced pulling by manpower is the only option, easily leading to equipment jamming, overturning, or damage. There is a lack of a coordinated control system that can intelligently adjust traction and autonomous power based on real-time resistance.

[0003] Poor environmental adaptability and lack of proactive obstacle-crossing capability: The existing support platform has fixed buoyancy and cannot adapt to changes in water level and silt depth within the pipeline, leading to frequent instances of cameras being submerged or the platform becoming stranded. More importantly, the current device completely lacks proactive obstacle-clearing capabilities. Once it encounters a substantial obstacle that cannot be bypassed, the entire inspection operation is interrupted, requiring expensive and time-consuming manual obstacle clearing, resulting in extremely low continuity and automation of the inspection process.

[0004] The detection process suffers from low intelligence and isolated data: Traditional CCTV inspection heavily relies on ground operators watching video in real time and manually recording defects. This method is inefficient, highly subjective, and prone to missed defects due to fatigue. Video data is isolated from other environmental parameters within the pipeline, making correlation analysis impossible. Current technology lacks an intelligent system capable of identifying defects in real time at the terminal and linking the identification results with equipment control.

[0005] The existing detection devices suffer from limited functionality and insufficient coordination: they typically only perform the core task of image acquisition, with modules for movement, shooting, positioning, and analysis operating independently. The lack of a central controller capable of deeply integrating and intelligently decision-making with the power system, buoyancy system, obstacle removal mechanism, and sensor data prevents the system from functioning as a cohesive intelligent entity to cope with the complex and ever-changing environment within pipelines, thus limiting further improvements in detection efficiency and depth. Therefore, a pipeline CCTV detection device and method for complex operating conditions needs to be designed to address these issues. Summary of the Invention

[0006] The purpose of this invention is to address the aforementioned problems by proposing a pipeline CCTV inspection device and method for complex working conditions. When pipelines or culverts and similar engineering pipelines are operating with water and have not been dredged, and inspection is required, the device is placed directly inside the pipeline for CCTV inspection. This pipeline CCTV inspection device and method integrates intelligent movement, adaptive buoyancy, active obstacle removal, multi-sensor fusion, and AI real-time recognition, and allows for deep collaboration and linkage between various systems. This fundamentally solves the problems of difficult inspection, low efficiency, shallow data, and high interruption rate under complex working conditions.

[0007] Therefore, the present invention employs the following technical methods: A pipeline CCTV inspection device for complex working conditions includes a hull platform carrying CCTV inspection instruments, and a power system consisting of an active propulsion system and a tunnel traction system. The active propulsion system is installed at the stern of the hull platform to provide autonomous power. The tunnel traction system includes a winch unit and a traction rope installed at the pipeline opening to provide traction power from outside the pipeline. The power system is electrically connected to an intelligent power controller, which coordinates the active propulsion system and the tunnel traction system to automatically distribute power output according to the resistance inside the pipeline. The hull platform is equipped with a variable buoyancy system that adjusts the water volume in the ballast tanks to change the draft by using water pumps; the front of the hull platform is equipped with a clearing mechanism to remove or break up obstacles in front of it. The upper surface of the hull platform is equipped with a multi-sensor system to collect data on the pipeline's attitude, environment, gas, and cross-section; the CCTV inspection instrument has a built-in AI image recognition module to analyze the CCTV video stream in real time and identify pipeline defects and blockages.

[0008] Preferably, the active propulsion system is a waterproof electric propeller with adjustable thrust direction; the winch unit of the tunnel traction system is equipped with a tension sensor and encoder for real-time measurement of traction force. and speed of travel The active propulsion system incorporates a current sensor to monitor the current of the ship's active propulsion motor. This current value is related to the thruster's output torque / thrust. Proportional.

[0009] Preferably, the intelligent power controller performs closed-loop control of traction force and travel speed based on PID control, and its specific calculation process includes: Speed ​​error is calculated using the speed control loop. The required total thrust is calculated using a PID controller. : ; ; in, For the proportional, integral, and derivative gain coefficients of the speed loop PID controller; Power distribution and drag monitoring, including: Real-time estimation of total resistance: The system estimates the total resistance in real time by measuring the traction force. and thruster current To estimate the current total resistance in the pipe : ; in K This is the thruster current-thrust conversion factor; Dynamically distributed thrust: On flat, low-resistance sections of road, the winch provides traction. Active thrusters are either in use or on standby. Resistance Adaptive: When estimating total resistance When it increases: The controller first instructs the winch to increase the traction force proportionally. ; like Reaching the preset safety limit The controller then activates the active thrusters to provide additional thrust. ,make sure ; Anti-jamming coordination: If the system detects... but If the jamming threshold is exceeded, the controller will execute a shaking strategy: Release the tether and simultaneously command the active thrusters to reverse at full power for a moment to get out of trouble, then try to move forward together again; Winch traction force setting: ; Thruster thrust setting: ; in R For the dynamic distribution ratio of traction force (0 <R ≤ 1), This is the maximum safe tension of the rope.

[0010] Preferably, the intelligent power controller is linked with the variable buoyancy system, the obstacle removal mechanism, and the multi-sensor system to execute a progressive obstacle response strategy, specifically including: When the real-time total resistance Continuously exceeding the first warning threshold Reaching the first time When, that is, satisfied At that time, the controller prioritizes instructing the variable buoyancy system to increase buoyancy; If at the second time Internal, average resistance The controller then enhances the power of the active propulsion system; If the power increases and the resistance continues to exceed the second barrier threshold... That is, satisfying and ,continued When this occurs, the controller instructs the obstacle removal mechanism to start; This represents the minimum creeping speed.

[0011] Preferably, the AI ​​image recognition module is built based on a convolutional neural network model, and its specific technical solution includes: Model input: Video frames of the inner wall of the pipe captured in real time by a CCTV camera; Model output: The recognition result of the target in the image, including the target category and confidence C; the target category includes structural defects and functional defects; structural defects include cracks, fractures, misalignments, and disconnections, and functional defects include deposits, scale, root intrusion, and obstacle blockage; Linkage control: The output of the AI ​​image recognition module is linked with the intelligent power controller and the obstacle removal mechanism.

[0012] Preferably, the linkage between the AI ​​image recognition module and the intelligent power controller includes specific calculations: When a structural defect is identified and its confidence level is... At that time, the controller automatically modifies the ship's speed setpoint to the fine-detection speed: ; ; k The deceleration coefficient is 0. <k<1; The confidence threshold for structural defects; When the obstacle blockage category in the functional defect is identified and the confidence level is... At that time, this information is used as a priority condition to trigger the progressive obstacle response strategy, and the estimated resistance at that time is immediately applied. With threshold and Compare; The threshold for obstacle confidence.

[0013] Preferably, the linkage between the AI ​​image recognition module and the obstacle removal agency includes specific location calculations: When an obstacle is detected, obtain the x-coordinate of the center point of its image bounding box. and frame width ; Convert image coordinates into deflection angle commands for the obstacle-clearing robotic arm: ; in, The midpoint angle of the robotic arm. The x-coordinate of the image center. This represents the maximum deflection range of the robotic arm.

[0014] Preferably, the multi-sensor system includes an inertial measurement unit (IMU), a laser scanner, a gas sensor, a sonar probe, and a water quality sensor; the intelligent power controller is also used to perform the following calculations: Cross-section passability judgment: Calculate the effective pipe diameter based on laser scanner data. The diameter of the pipe as measured by laser. The depth of silt estimated by sonar or laser; when At that time, it was determined that the lifting passage was not supported; To ensure the safe passage of the ship through the diameter; Attitude compensation: When the obstacle clearance mechanism moves or experiences a sudden change in power, the roll angle is measured based on the IMU. and pitch angle Calculate the attitude compensation amount: ; It is differentially applied to the left and right thrusters to generate anti-rollover torque.

[0015] Preferably, the variable buoyancy system performs the following closed-loop control based on the silt depth data fed back from the laser rangefinder or sonar in the multi-sensor system: Calculate the difference between the camera target height and the current silt surface height: ; Calculate the target displacement of the ballast tanks using a PI controller: ; Control the water pump operation to make the actual displacement of the ballast tank approach that of the target volume. This is to keep the CCTV inspection equipment's camera at the optimal shooting height.

[0016] Preferably, the detection method of the pipeline CCTV inspection device under the above-mentioned complex working conditions includes the following steps: S1, to cut off the flow in the pipeline, pump water and ventilate it; S2, install the wellhead traction system and thread the traction rope to the upstream and downstream wellheads; S3, lower the hull platform and CCTV inspection instruments to the bottom of the well, and start the multi-sensor system and AI image recognition module; S4 controls the hull's movement within the pipeline through a smart power controller that coordinates with the self-propulsion and traction system. S5 utilizes an AI image recognition module to analyze the video stream in real time. When a structural defect is detected, it automatically slows down to perform detailed shooting. When an obstruction is detected, it triggers the progressive obstacle response strategy. After completing the inspection, the S6 retrieves the equipment and outputs a comprehensive inspection report that integrates sensor data, geographic information, and AI recognition results.

[0017] The beneficial effects of this invention are as follows: 1. This invention achieves stable, controllable movement and accurate detection under complex working conditions, fundamentally overcoming the shortcomings of insufficient power and coarse control in traditional equipment. Through an intelligent power controller and closed-loop control algorithm, the traction force at the tunnel entrance and the self-propulsion force of the hull are deeply integrated. The system can estimate the resistance in real time and dynamically allocate power output, ensuring sufficient traction in silt to prevent slippage, while avoiding power waste and speed fluctuations in smooth pipe sections. When a structural defect is detected, the system can automatically reduce speed to a fine detection mode to achieve uniform speed or hovering shooting, ensuring clear, stable images and complete data. This solves the fundamental problems of uneven traction force, wheel slippage, and inability to adaptively adjust speed according to the detection content in traditional manual methods, upgrading movement control from relying on human experience to being data-driven.

[0018] 2. This invention constructs a systematic, progressive obstacle-handling strategy, significantly improving the continuity and automation of inspection operations and effectively reducing the risk of operational interruptions. Addressing the pain point of existing devices stopping immediately upon encountering an obstacle, it establishes an intelligent response mechanism of "flexible avoidance - power enhancement - active removal." The system prioritizes attempting to lift the hull and maneuver around soft obstacles using a variable buoyancy system to minimize energy consumption. If this fails, it coordinates enhanced traction and propulsion power to attempt forced passage. For substantially hard obstacles, it activates the obstacle-clearing mechanism for localization and removal. This multi-level, orderly response strategy endows the device with obstacle-crossing and obstacle-clearing capabilities not found in traditional equipment, minimizing the frequency of manual intervention and operational interruptions caused by obstacles during inspection, thereby ensuring the efficient and continuous execution of long-distance, complex pipeline inspection tasks.

[0019] 3. This invention achieves a leap from passive recording to proactive perception and intelligent decision-making, generating a deeply integrated, multi-dimensional pipeline health profile. By linking AI image recognition, multi-sensor data, and the core control system, the AI ​​module can not only identify defect categories in real time but also directly trigger control commands, giving the detection equipment preliminary on-site decision-making capabilities. Simultaneously, all identified defect information, environmental data, and geometric measurement data are synchronously tagged with location and fused together, ultimately generating a comprehensive diagnostic report integrating visual, environmental, and structural information. This completely changes the outdated model of relying on manual video interpretation and isolated data in traditional methods, providing users with a more comprehensive, objective, and in-depth assessment of pipeline health status, greatly enhancing the value of detection results and the scientific rigor of subsequent repair plan development. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the hull platform and load-bearing testing instrument of the present invention; Figure 2 This is a block diagram showing the connection logic of the device of the present invention; Figure 3 This is a schematic diagram of the method flow of the present invention; Figure 4 This is a schematic diagram illustrating the implementation of a progressive obstacle response strategy in an embodiment of the present invention; Figure 5 This is a schematic diagram of a CCTV detection instrument in an embodiment of the present invention; The attached diagram is labeled as follows: hull platform 1, towing rope 2, CCTV inspection instrument 3. Detailed Implementation

[0021] Example 1: like Figures 1-2 A pipeline CCTV inspection device for complex working conditions includes a hull platform 1 that carries the CCTV inspection instrument 3, and a power system consisting of an active propulsion system and a tunnel traction system. The active propulsion system is installed at the stern of the hull platform 1 to provide autonomous power. The tunnel traction system includes a winch unit and a traction rope 2 installed at the pipeline opening to provide traction power from outside the pipeline. The power system is electrically connected to an intelligent power controller, which coordinates the active propulsion system and the tunnel traction system to automatically distribute power output according to the resistance inside the pipeline. The hull platform 1 is equipped with a variable buoyancy system, which adjusts the water volume in the ballast tank to change the draft by means of a water pump; the front of the hull platform 1 is equipped with a clearing mechanism to clear or break obstacles in front; the upper surface of the hull platform 1 is equipped with a multi-sensor system to collect data on the pipeline's attitude, environment, gas, and cross-section; the CCTV inspection instrument 3 has a built-in AI image recognition module to analyze the CCTV video stream in real time and identify pipeline defects and blockages.

[0022] Preferably, the active propulsion system is a waterproof electric propeller with adjustable thrust direction; the winch unit of the tunnel traction system is equipped with a tension sensor and encoder for real-time measurement of traction force. and speed of travel The active propulsion system incorporates a current sensor to monitor the current of the ship's active propulsion motor. This current value is related to the thruster's output torque / thrust. Proportional.

[0023] Preferably, the intelligent power controller performs closed-loop control of traction force and travel speed based on PID control, and its specific calculation process includes: Speed ​​error is calculated using the speed control loop. The required total thrust is calculated using a PID controller. : ; ; in, For the proportional, integral, and derivative gain coefficients of the speed loop PID controller; Power distribution and drag monitoring, including: Real-time estimation of total resistance: The system estimates the total resistance in real time by measuring the traction force. and thruster current To estimate the current total resistance in the pipe : ; in K This is the thruster current-thrust conversion factor; Dynamically distributed thrust: On flat, low-resistance sections of road, the winch provides traction. Active thrusters are either in use or on standby. Resistance Adaptive: When estimating total resistance When it increases: The controller first instructs the winch to increase the traction force proportionally. ; like Reaching the preset safety limit The controller then activates the active thrusters to provide additional thrust. ,make sure ; Anti-jamming coordination: If the system detects... but If the jamming threshold is exceeded, the controller will execute a shaking strategy: Release the tether and simultaneously command the active thrusters to reverse at full power for a moment to get out of trouble, then try to move forward together again; Winch traction force setting: ; Thruster thrust setting: ; in R For the dynamic distribution ratio of traction force (0 <R ≤ 1), This is the maximum safe tension of the rope.

[0024] like Figure 4 As shown, preferably, the intelligent power controller is linked with the variable buoyancy system, the obstacle removal mechanism, and the multi-sensor system to execute a progressive obstacle response strategy, specifically including: When the real-time total resistance Continuously exceeding the first warning threshold Reaching the first time When, that is, satisfied At that time, the controller prioritizes instructing the variable buoyancy system to increase buoyancy; If at the second time Internal, average resistance The controller then enhances the power of the active propulsion system; If the power increases and the resistance continues to exceed the second barrier threshold... That is, satisfying and ,continued When this occurs, the controller instructs the obstacle removal mechanism to start; This represents the minimum creeping speed.

[0025] Preferably, the AI ​​image recognition module is built based on a convolutional neural network model, and its specific technical solution includes: Model input: Video frames of the inner wall of the pipe captured in real time by a CCTV camera; Model output: The recognition result of the target in the image, including the target category and confidence C; the target category includes structural defects and functional defects; structural defects include cracks, fractures, misalignments, and disconnections, and functional defects include deposits, scale, root intrusion, and obstacle blockage; Linkage control: The output of the AI ​​image recognition module is linked with the intelligent power controller and the obstacle removal mechanism.

[0026] Preferably, the linkage between the AI ​​image recognition module and the intelligent power controller includes specific calculations: When a structural defect is identified and its confidence level is... At that time, the controller automatically modifies the ship's speed setpoint to the fine-detection speed: ; ; k The deceleration coefficient is 0. <k<1; The confidence threshold for structural defects; When the obstacle blockage category in the functional defect is identified and the confidence level is... At that time, this information is used as a priority condition to trigger the progressive obstacle response strategy, and the estimated resistance at that time is immediately applied. With threshold and Compare; The threshold for obstacle confidence.

[0027] Preferably, the linkage between the AI ​​image recognition module and the obstacle removal agency includes specific location calculations: When an obstacle is detected, obtain the x-coordinate of the center point of its image bounding box. and frame width ; Convert image coordinates into deflection angle commands for the obstacle-clearing robotic arm: ; in, The midpoint angle of the robotic arm. The x-coordinate of the image center. This represents the maximum deflection range of the robotic arm.

[0028] Preferably, the multi-sensor system includes an inertial measurement unit (IMU), a laser scanner, a gas sensor, a sonar probe, and a water quality sensor; the intelligent power controller is also used to perform the following calculations: Cross-section passability judgment: Calculate the effective pipe diameter based on laser scanner data. The diameter of the pipe as measured by laser. The depth of silt estimated by sonar or laser; when At that time, it was determined that the lifting passage was not supported; To ensure the safe passage of the ship through the diameter; Attitude compensation: When the obstacle clearance mechanism moves or experiences a sudden change in power, the roll angle is measured based on the IMU. and pitch angle Calculate the attitude compensation amount: ; It is differentially applied to the left and right thrusters to generate anti-rollover torque.

[0029] Preferably, the variable buoyancy system performs the following closed-loop control based on the silt depth data fed back from the laser rangefinder or sonar in the multi-sensor system: Calculate the difference between the camera target height and the current silt surface height: ; Calculate the target displacement of the ballast tanks using a PI controller: ; Control the water pump operation to make the actual displacement of the ballast tank approach that of the target volume. This is to keep the camera of CCTV inspection instrument 3 at the optimal shooting height.

[0030] Furthermore, the active propulsion system is a waterproof electric propeller with adjustable thrust direction and powered by an onboard battery.

[0031] Furthermore, the obstacle removal mechanism is a rotatable hydraulic arm equipped with a mechanical claw or cutter at its end, and its activation is triggered by the AI ​​image recognition module after it detects an obstruction.

[0032] Furthermore, such as Figure 5 As shown, the CCTV inspection instrument is a conventional CCTV inspection instrument that has been disassembled from the wheels and mounted on the ship's platform.

[0033] Example 2: like Figure 3 As shown, the detection method of the pipeline CCTV inspection device under the above-mentioned complex working conditions includes the following steps: S1, to cut off the flow in the pipeline, pump water and ventilate it; S2, Install the wellhead traction system and thread traction rope 2 to the upstream and downstream wellheads; S3, lower the hull platform 1 and CCTV detection instrument 3 to the bottom of the well, and start the multi-sensor system and AI image recognition module; S4 controls the hull's movement within the pipeline through a smart power controller that coordinates with the self-propulsion and traction system. S5 utilizes an AI image recognition module to analyze the video stream in real time. When a structural defect is detected, it automatically slows down to perform detailed shooting. When an obstruction is detected, it triggers the progressive obstacle response strategy. After completing the inspection, the S6 retrieves the equipment and outputs a comprehensive inspection report that integrates sensor data, geographic information, and AI recognition results.

Claims

1. A pipeline CCTV inspection device for complex working conditions, comprising a ship platform (1) supporting CCTV inspection instruments (3), characterized in that, It also includes a power system consisting of an active propulsion system and a tunnel traction system. The active propulsion system is installed at the stern of the hull platform (1) to provide autonomous power. The tunnel traction system includes a winch unit and a traction rope (2) installed at the pipe opening to provide traction power from outside the pipe. The power system is electrically connected to an intelligent power controller. The intelligent power controller coordinates the active propulsion system and the tunnel traction system and automatically distributes power output according to the resistance inside the pipe. The hull platform (1) is equipped with a variable buoyancy system, which adjusts the water volume of the ballast tank through a water pump to change the draft. The front end of the hull platform (1) is equipped with a clearing mechanism to clear or break obstacles in front. The upper surface of the hull platform (1) is equipped with a multi-sensor system to collect data on the attitude, environment, gas and cross-section inside the pipe. The CCTV detection instrument (3) has a built-in AI image recognition module to analyze the CCTV video stream in real time and identify pipe defects and blockages.

2. The pipeline CCTV inspection device under complex working conditions according to claim 1, characterized in that: The active propulsion system is a waterproof electric propeller with adjustable thrust direction; the winch unit of the tunnel traction system is equipped with a tension sensor and encoder for real-time measurement of traction force. and speed of travel ; The active propulsion system incorporates a current sensor to monitor the current of the ship's active propulsion motor. This current value is related to the thruster's output torque / thrust. Proportional.

3. The pipeline CCTV inspection device under complex working conditions according to claim 2, characterized in that: The intelligent power controller performs closed-loop control of traction force and travel speed based on PID control, and its specific calculation process includes: Speed ​​error is calculated using the speed control loop. The required total thrust is calculated using a PID controller. ; Power distribution and drag monitoring, including: Real-time estimation of total resistance: The system estimates the total resistance in real time by measuring the traction force. and thruster current To estimate the current total resistance in the pipe ; Dynamically distributed thrust: On flat, low-resistance sections of road, the winch provides traction. Active thrusters are either in use or on standby. Resistance adaptive: When estimating total resistance When the force is increased, the controller first instructs the winch to increase the traction force proportionally. ; like Reaching the preset safety limit The controller then activates the active thrusters to provide additional thrust. ,make sure ; Anti-jamming coordination: If the system detects... but If the jamming threshold is exceeded, the controller executes a swaying strategy: loosens the traction rope and simultaneously commands the active thrusters to reverse at full force for a moment to get out of trouble, and then tries to move forward together again.

4. The pipeline CCTV inspection device under complex working conditions according to claim 3, characterized in that: The intelligent power controller, in conjunction with the variable buoyancy system, obstacle removal mechanism, and multi-sensor system, executes a progressive obstacle response strategy, specifically including: When the real-time total resistance Continuously exceeding the first warning threshold Reaching the first time When, that is, satisfied At that time, the controller prioritizes instructing the variable buoyancy system to increase buoyancy; If at the second time Internal, average resistance The controller then enhances the power of the active propulsion system; If the power increases and the resistance continues to exceed the second barrier threshold... That is, satisfying and ,continued When this occurs, the controller instructs the obstacle removal mechanism to start; This represents the minimum creeping speed.

5. The pipeline CCTV inspection device under complex working conditions according to claim 1, characterized in that: The AI ​​image recognition module is built on a convolutional neural network model, and its specific technical solution includes: Model input: Video frames of the inner wall of the pipe captured in real time by a CCTV camera; Model output: The recognition result of the target in the image, including the target category and confidence C; the target category includes structural defects and functional defects; structural defects include cracks, fractures, misalignments, and disconnections, and functional defects include deposits, scale, root intrusion, and obstacle blockage; Linkage control: The output of the AI ​​image recognition module is linked with the intelligent power controller and the obstacle removal mechanism.

6. The pipeline CCTV inspection device under complex working conditions according to claim 5, characterized in that: The linkage between the AI ​​image recognition module and the intelligent power controller includes specific calculations: When a structural defect is identified and its confidence level is... At that time, the controller automatically modifies the ship's speed setpoint to the fine-detection speed: ； ； k is the deceleration coefficient, 0 < k < 1; The confidence threshold for structural defects; When the obstacle blockage category in the functional defect is identified and the confidence level is... At that time, this information is used as a priority condition to trigger the progressive obstacle response strategy, and the estimated resistance at that time is immediately applied. With threshold and Compare; The threshold for obstacle confidence.

7. The pipeline CCTV inspection device under complex working conditions according to claim 5, characterized in that: The linkage between the AI ​​image recognition module and the obstacle removal agency includes specific location calculations: When an obstacle is detected, obtain the x-coordinate of the center point of its image bounding box. and frame width ; Convert image coordinates into deflection angle commands for the obstacle-clearing robotic arm: ； in, The midpoint angle of the robotic arm. The x-coordinate of the image center. This represents the maximum deflection range of the robotic arm.

8. The pipeline CCTV inspection device under complex working conditions according to claim 1, characterized in that: The multi-sensor system includes an inertial measurement unit (IMU), a laser scanner, a gas sensor, a sonar probe, and a water quality sensor; the intelligent power controller is also used to perform the following calculations: Cross-section passability judgment: Calculate the effective pipe diameter based on laser scanner data. The diameter of the pipe as measured by laser. The depth of silt estimated by sonar or laser; when At that time, it was determined that the lifting passage was not supported; To ensure the ship's safe passage through the diameter; Attitude compensation: When the obstacle clearance mechanism moves or experiences a sudden change in power, the roll angle is measured based on the IMU. and pitch angle Calculate the attitude compensation amount: ； It is differentially applied to the left and right thrusters to generate anti-rollover torque.

9. A pipeline CCTV inspection device under complex working conditions according to claim 1, characterized in that: The variable buoyancy system performs the following closed-loop control based on the silt depth data fed back by the laser rangefinder or sonar in the multi-sensor system: Calculate the difference between the camera target height and the current silt surface height: ； Calculate the target displacement of the ballast tanks using a PI controller: ； Control the water pump operation to make the actual displacement of the ballast tank approach that of the target volume. In order to keep the camera of the CCTV inspection instrument (3) at the optimal shooting height.

10. The detection method of a pipeline CCTV inspection device under complex working conditions according to any one of claims 1-9, characterized in that, Includes the following steps: S1, to cut off the flow in the pipeline, pump water and ventilate it; S2, install the traction system at the wellhead and thread the traction rope (2) to the upstream and downstream wellheads; S3, place the hull platform (1) and CCTV detection instrument (3) into the bottom of the well, and start the multi-sensor system and AI image recognition module; S4 controls the hull's movement within the pipeline through a smart power controller that coordinates with the self-propulsion and traction system. S5 utilizes an AI image recognition module to analyze the video stream in real time. When a structural defect is detected, it automatically slows down to perform detailed shooting. When an obstruction is detected, it triggers the progressive obstacle response strategy. After completing the inspection, the S6 retrieves the equipment and outputs a comprehensive inspection report that integrates sensor data, geographic information, and AI recognition results.

Images