Visual inspection robot system and method for small-caliber pipeline inspection

Abstract

The invention relates to the field of pipeline robots, in particular to a visual inspection robot system and method for small-caliber pipeline inspection. The pipeline detection robot system comprises the following structures: a crawler-type driving unit, a liftable holder, a visual information acquisition unit and a controller unit. The detection method comprises the following steps that after the pipeline robot confirms that the function is normal, a user issues a voice instruction through a remote terminal, voice recognition software recognizes voice into characters, and a Qwen-4B large model converts the character instruction into an ROS system executable instruction; after the controller unit receives the instruction, related data of the pipeline are collected through multiple sensors, the data are transmitted to a remote terminal in real time to be displayed, and meanwhile the data are stored in a controller; and after the detection is completed, the user sends a detection ending instruction to the controller to enable the controller to complete equipment resetting, and data is locally stored. The method has the advantages of being high in detection comprehensiveness, high in maneuverability and flexibility, high in precision and reliability and the like.

Description

Technical Field

[0001] This invention relates to the field of pipeline robots, and in particular to a vision inspection robot system and method for inspecting small-diameter pipelines. Background Technology

[0002] Pipeline systems are a major infrastructure for modern industrial production, transportation, and municipal services. They involve energy transmission, sewage discharge, gas supply, and many other aspects, and their operation is vital to industrial production, urban functions, and people's lives. However, during long-term service, pipelines are affected by various factors such as underground soil corrosion, media erosion, hidden dangers left over from construction, changes in temperature stress, and external impacts. These factors lead to various types of defects in the pipeline, such as internal cracks, thinning of the pipe wall due to corrosion, blockage and accumulation of foreign objects, and structural deformation and misalignment. These defects are initially hidden, and if the problem is not quickly and accurately located and repaired, they may lead to a series of serious consequences over time. For example, blocked municipal drainage pipes can cause urban flooding, causing great inconvenience to people's daily travel; leaks in industrial oil and gas pipelines can easily lead to accidents such as fires or explosions, causing huge property damage; and damage to water supply pipelines can cause precious water resources to be wasted, and even pollute groundwater, directly affecting the health of the general public. Therefore, developing a comprehensive and highly efficient method for detecting and repairing pipeline defects has become a very important task.

[0003] Reference 1 (Chinese Patent Application No.: CN202511237589.1) discloses an adaptive quadrupedal bionic robot for overcoming obstacles in pipe walls. It utilizes information obtained from 3D vision sensors, LiDAR, inertial measurement units, and joint torque sensors to control the system's operation, accurately detecting obstacles and autonomously planning its gait for overcoming them. However, this mechanism is too bulky and may not be suitable for small-diameter pipes. Reference 2 (Zhao Xingliang. Research on Defect Detection and Operation System of Drainage Pipe Clearing Robot [D]. Beijing University of Civil Engineering and Architecture, 2024) proposes a drainage pipe clearing robot system for pipe diameters of 300-800mm. This system uses hardware structure, defect detection model, positioning method, and operation planning to complete the task of clearing hard defects. However, this system does not provide natural language voice control and relies solely on manual operation, resulting in low efficiency. Furthermore, it does not incorporate other types of sensors such as lasers and acoustics, therefore defect detection is not comprehensive enough, its adaptability to complex pipe environments needs improvement, and image clarity also requires enhancement.

[0004] Current technologies have many shortcomings, which limits the practical application of pipeline inspection. Manual operation is inadequate for long-distance, complex inspections, being both inefficient and dangerous. A single inspection dimension and limited environmental adaptability can easily lead to missed or false defects, impacting maintenance decisions. Image acquisition quality is affected by environmental factors such as dim lighting and uneven illumination, further limiting defect accuracy. Therefore, there is an urgent need for a new pipeline robot visual information acquisition unit and method. Utilizing natural language interaction technology, this would allow users to directly issue commands for complex inspection tasks, automatically supplementing command information through contextual inference from a large model, simplifying operation steps. Combining a multi-angle high-definition visual information acquisition unit with various types of sensors, including laser, acoustic, and ultrasonic sensors, would achieve comprehensive and accurate detection of various defect types, improving inspection efficiency, reducing maintenance risks, and further expanding its application in small-diameter, complex pipelines, providing more reliable protection for the safe operation and maintenance of pipeline infrastructure. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a vision inspection robot system and method for small-diameter pipe inspection using a large model, which solves the problems of pipe robots struggling to detect various pipe defects and being unable to flexibly control camera image capture.

[0006] To achieve this objective, the present invention adopts the following technical solution: A voice-controlled visual information acquisition unit for a pipeline robot with a large-scale application model includes the following structure: frame, power supply module, tracked drive unit, sensor module, controller unit, lighting module, and remote terminal module; The power supply module is located in the middle of the vehicle frame; The tracked drive unit is symmetrically arranged on both sides of the vehicle frame; The sensor modules are located at multiple positions on the vehicle frame; The lifting gimbal module is located at the rear end of the gyroscope module; The visual information acquisition unit is located at the upper end of the lifting platform; The controller unit is located at the rear end of the lifting gimbal module.

[0007] The lighting module is located at the rear of the vehicle frame; Furthermore, the power supply module is electrically connected to the pipeline robot via a wired cable interface to supply power to the tracked drive unit, lifting gimbal module, visual information acquisition unit, gyroscope module, laser module, acoustic module, lighting module, and controller unit.

[0008] Furthermore, the tracked drive unit includes a bearing, a motor, a track, a drive wheel, and a driven wheel. The motor is mounted on the frame, and the output shaft of the motor is coupled to the bearing. The tracked drive unit carries the pipeline robot to move.

[0009] Furthermore, the sensor module includes a gyroscope sensor, a laser sensor, an acoustic sensor, and an ultrasonic sensor. The gyroscope sensor is installed at the rear end of the lifting gimbal module, providing positioning and attitude measurement functions for the pipeline robot. The laser sensor is located at the front right side of the frame, providing laser detection functions for the pipeline robot. The ultrasonic sensor is installed at the front right side of the frame, and the acoustic sensor is installed on the upper part of the visual information acquisition unit. The ultrasonic sensor and the acoustic sensor provide sound wave detection functions for the pipeline robot.

[0010] Furthermore, the lifting gimbal module includes a rotating base, a fixed housing, and connecting rods. The fixed housing is fixedly connected to the upper surface of the rotating base, and its side wall has pre-set coaxial connection holes adapted to the connecting rods. The connecting rods include two sets of parallelogram connecting rod assemblies, each set of assemblies containing two long connecting rods of the same length. One end of the long connecting rod is hinged to a hole in the side wall of the fixed housing, and the other end is used to connect to the visual information acquisition unit.

[0011] Furthermore, the visual information acquisition unit includes a camera, a camera mounting bracket, a rotating platform, a camera platform, an auxiliary ring light, an illumination platform, a housing connecting plate, a supporting housing, and hinge pins. The hinge pins are mounted on both sides of the supporting housing and hinged to the lifting gimbal module. The housing connecting plate is fixedly connected to the supporting housing and also to the illumination platform. The auxiliary ring light is mounted on the illumination platform, the camera platform is mounted on the illumination platform, the rotating platform is mounted on the camera platform, and the camera is secured to the camera platform via the camera mounting bracket, while also being hinged to the camera mounting bracket.

[0012] Furthermore, the lighting module includes a searchlight, which is divided into two parts: one part is connected to the front cover of the vehicle frame, and the other part is installed at the front end of the upper gimbal and electrically connected to the control module. It is used to provide a light source in the dark pipe to facilitate camera sampling.

[0013] Furthermore, the controller unit is constructed from a Raspberry Pi 5 development board, which deploys the Qwen-4B large language model and the ROS system. Upon receiving natural language commands from the remote terminal, the development board processes them into machine language using the large language model. The ROS system then receives the machine language and distributes control signals to each module, thereby controlling the various modules of the robot. The control module is electrically connected to the tracked drive unit, the lifting gimbal module, the visual information acquisition unit, and the sensor module. It can control the motor speed and track steering angle of the tracked drive unit, adjust the rotation degree of the robotic arm of the lifting gimbal module, adjust the camera angle of the visual information acquisition unit, and adjust the brightness of the spotlight in the lighting module. Simultaneously, it can transmit the motor speed and track angle status of the tracked drive unit, the images acquired by the visual information acquisition unit, the position information of the gyroscope module, the height of the lifting gimbal, the brightness of the spotlight, and the information from the laser sensor, acoustic sensor, and ultrasonic sensor back to the remote terminal for display.

[0014] Furthermore, the remote terminal module consists of a computer, host computer software, and a microphone. It communicates with the control module via a wired connection. The user transmits voice commands to the host computer software through the microphone. The host computer software transmits the user's natural language voice commands to the development board via wired communication. At the same time, it receives image information from the visual information acquisition unit and status information of the pipeline robot.

[0015] The above technical solution has the following beneficial effects: 1. It has a wide defect detection coverage, integrating multiple detection modules such as image acquisition, laser, acoustic, and ultrasonic. It can detect pipelines from multiple angles such as vision, optics, and acoustics, and can identify various types of pipeline defects such as cracks, corrosion, deformation, and internal looseness. It overcomes the problem that single detection methods are difficult to detect complex defects, and greatly improves the coverage and comprehensiveness of pipeline defect detection.

[0016] 2. The pipeline robot is highly maneuverable. The tracked drive unit allows the system to move stably inside the pipeline, even when the inner wall is uneven or there are local obstacles. The lifting gimbal module can flexibly adjust the spatial position and angle of image acquisition or other detection modules, enabling comprehensive detection of the top, sides and bottom of the pipeline, greatly improving the system's adaptability to complex pipeline environments.

[0017] 3. The collaborative control is highly efficient. Users can directly issue complex tasks using natural language. The large model can understand the general meaning of the instructions and can make inferences based on the context. It can perform real-time analysis and processing of data from multiple sensors, and integrate visual, auditory, and vibration data to greatly improve the accuracy of defect identification.

[0018] 4. Good detection accuracy and reliability: The technologies of imaging, laser, acoustics, and ultrasound work together and corroborate each other. On the one hand, images provide direct visual characteristics of defects, while lasers provide precise measurement, ultrasound delves into internal defects, and acoustics analyzes abnormal vibrations. This multi-faceted detection of defects and their detailed conditions reduces the occurrence of omissions or misjudgments, and significantly improves the accuracy and reliability of pipeline defect detection.

[0019] 5. Reduced manpower and material resources: The cost of image acquisition is lower, and there is no need for professional staff to constantly adjust instructions based on sensor information. The large language model will automatically adjust the current status of each module of the robot based on sensor information, which greatly reduces labor costs. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall structure of the visual information acquisition unit for the pipeline robot of the present invention.

[0021] Figure 2 This is a schematic diagram of the tracked drive unit structure of the visual information acquisition unit of the pipeline robot of the present invention.

[0022] Figure 3 This is a schematic diagram of the lifting gimbal module structure of the visual information acquisition unit for the pipeline robot of the present invention.

[0023] Figure 4 This is a schematic diagram of the visual information acquisition unit and lighting module of the visual information acquisition unit of the pipeline robot of the present invention.

[0024] Figure 5 This is a schematic diagram of the communication structure of the visual information acquisition unit of the pipeline robot of the present invention.

[0025] Figure 6 This is the host computer interface for the visual information acquisition unit of the pipeline robot of the present invention. Detailed Implementation

[0026] Specific embodiments of the invention are described in detail below. These examples are illustrated in the accompanying drawings, and throughout the description, the same or similar reference numerals represent components that have the same or similar functions. The examples set forth below are intended to provide a perspective for understanding the invention; they are illustrative and not restrictive, and are only used to illustrate the core concepts and scope of application of the invention.

[0027] In describing this invention, it should be understood that directional terms such as "longitudinal," "lateral," "up," "down," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer" refer to positional relationships based on the accompanying drawings. These terms are intended to facilitate a clear description of the invention and simplify the explanatory process, and do not mandate that actual devices or components must be designed, manufactured, or operated according to these specific orientations. Therefore, they should not be misunderstood as constraints on the scope of this invention. Features marked "first" and "second" may explicitly or implicitly include one or more of these features, and their purpose is to distinguish the objects being described, without involving any order or priority of importance.

[0028] In the context of this invention, unless otherwise specified, the terms "multiple" or "several" refer to two or more quantities.

[0029] In describing this invention, it should be clearly stated that, unless otherwise specified and limited, the terms "installation," "connection," and "linking" should have a broad scope of interpretation and are not limited to their inherent definitions. This includes, but is not limited to, permanent and detachable connections, physical bonding and electrical coupling, direct contact and indirect connections achieved through intermediaries, and connectivity involving internal channels or paths between components. Those skilled in the art will be able to understand the precise meaning of these terms in this invention according to the specific circumstances.

[0030] Figure 1 This is a schematic diagram of the overall structure of the pipeline inspection robot inspection system of the present invention. As shown in the figure, it includes the following structures: frame 101, power supply module 102, tracked drive unit 103, gyroscope sensor 104, lifting gimbal module 105, visual information acquisition unit 106, laser sensor 107, acoustic sensor 108, controller unit 109, and lighting module 110. The power supply module 102 is located in the middle of the vehicle frame 101; The tracked drive unit 103 is symmetrically arranged on both sides of the vehicle frame 101; The gyroscope sensor 104 is located at the front end of the vehicle frame 101; The lifting gimbal module 105 is located at the rear end of the gyroscope 104; The visual information acquisition unit 106 is located at the upper end of the lifting gimbal module 105; The laser sensor 107 is located at the left rear end of the frame 101; The ultrasonic sensor 108 is located at the right rear end of the frame 101; The controller unit 109 is located at the rear end of the lifting gimbal module 105.

[0031] The lighting module 110 is located at the rear end of the frame 101; The frame 101 serves as the base for the pipeline robot inspection system, comprising a base, a control unit, and various sensor modules. The base has shaft holes for fixing and securing with other modules.

[0032] As shown in the figure, the gyroscope sensor 104 is mounted on the frame 101. During the operation of the pipeline robot, the gyroscope module provides the controller with the position and attitude information of the pipeline robot by outputting angular velocity data in real time.

[0033] As shown in the figure, the laser module 107 includes a laser sensor that emits laser signals to assist the visual information acquisition unit and the acoustic module in performing three-dimensional contour scanning or distance measurement of the inner wall of the pipe.

[0034] As shown in the figure, the ultrasonic sensor 108 is mounted on the right front end of the chassis. The ultrasonic sensor emits ultrasonic waves towards the pipe wall. When the ultrasonic waves encounter defects inside the pipe or on the pipe wall during propagation, they generate reflected waves. The ultrasonic sensor receives the reflected waves, converts them into electrical signals, and transmits them to the controller unit. During the operation of the pipeline robot, the ultrasonic sensor 108 provides the pipeline robot with the function of detecting defects in the inner wall of the pipe, and can accurately locate the defect location.

[0035] As shown in the figure, the lighting module 110 includes a camera spotlight, which is electrically connected to the control module. During the pipeline robot's exploration process, the lighting module 110 provides a light source in the dark pipeline, facilitating sampling by the visual information acquisition unit 106.

[0036] The controller unit 109 includes a controller Raspberry Pi 5 development board, a motor drive module, and a wired communication module. The controller is mounted in the middle of the vehicle frame 101, and the wired communication module is located on the controller.

[0037] During the pipeline robot's exploration process, the Raspberry Pi 5 development board, after receiving natural language commands from the remote terminal, processes them into machine language through a large language model. The ROS system then receives the machine language and distributes control signals to each module, thereby controlling each module of the robot. The control module 109 is electrically connected to the tracked drive unit 103, the lifting gimbal module 105, the visual information acquisition unit 106, the gyroscope and attitude measurement module 104, the laser module 107, the acoustic module 108, and the lighting module 110. It can control the motor speed and track steering angle of the tracked drive unit 103, adjust the rotation degree of the robotic arm of the lifting gimbal module 105, adjust the camera angle of the visual information acquisition unit 106, and adjust the brightness of the searchlight of the lighting module 110. At the same time, it can transmit the motor speed and track angle status of the tracked drive unit 103, the images acquired by the visual information acquisition unit 106, the position information of the gyroscope and attitude measurement module 104, the height of the lifting gimbal module 105, the brightness of the lighting module 110, and the sensor information of the laser module 107 and the acoustic module 108 back to the remote terminal for display.

[0038] Figure 2 This is a schematic diagram of the tracked drive unit 103 of the pipeline robot inspection system of the present invention. As shown in the figure, the tracked drive unit 103 of the pipeline robot includes: a drive wheel 201, a driven wheel 202, a motor 203, and a track 204. When the pipeline robot is running, the controller unit 109 sends a signal to the motor 203. After receiving the signal, the motor rotates at a specified speed, driving the drive wheel 201 to rotate, thereby driving the track forward. When the controller unit 109 sends a steering signal, the motors on both sides adjust their rotation speed according to the control signal, thereby achieving a differential steering effect. This tracked drive structure realizes the functions of the pipeline robot's movement and steering.

[0039] Figure 3 This is a schematic diagram of the lifting platform module 105 of the pipeline robot inspection system of the present invention. As shown in the figure, it includes the following structures: a rotating base 301, a fixed housing 302, and a connecting rod 303. The rotating base 301 can rotate 360° around its central vertical axis. The fixed housing 302 is fixedly connected to the upper surface of the rotating base 301, and its side wall is pre-set with coaxial connection holes adapted to the connecting rod 303. The connecting rod 303 includes two sets of parallelogram connecting rod assemblies. Each set of assemblies includes two long connecting rods of the same length. One end of the long connecting rod is hinged to the hole on the side wall of the fixed housing 302, and the other end is used to connect to the visual information acquisition unit 106.

[0040] During the operation of the pipeline robot, the lifting gimbal module 105 provides a mounting platform for the pipeline robot that can freely change its height and angle, assisting the visual information acquisition unit 106 in acquiring images. The lifting gimbal module 105 can receive control signals from the controller unit 109, and its horizontal working position can be adjusted by the horizontal rotation of the rotating base 301. The connecting rod 303 can drive the visual information acquisition unit 106 to achieve vertical lifting and lowering adjustment. The rotating base 301 and the connecting rod 303 work together to achieve freedom in image acquisition, enabling the pipeline robot to obtain pipeline defect images from different angles.

[0041] Figure 4 This is a schematic diagram of the structure of the visual information acquisition unit 106 for the pipeline robot of the present invention.

[0042] As shown in the figure, the visual information acquisition unit 106 includes a camera 401, a camera mounting bracket 402, a rotating platform 403, a camera platform 404, an auxiliary ring light 405, a lighting platform 406, a housing connecting plate 407, a supporting housing 408, a hinge pin 409, and an acoustic sensor 410. The hinge pin 409 is mounted on both sides of the supporting housing 408 and hinged to the lifting gimbal module 105. The housing connecting plate 407 is fixedly connected to the supporting housing 408 and also to the lighting platform 406. The auxiliary ring light 405 is mounted on the lighting platform 406. The camera platform 404 is mounted on the lighting platform 406. The rotating platform 403 is mounted on the camera platform 404. The camera 401 is fastened to the camera platform 404 via the camera mounting bracket 402, and is also hinged to the camera mounting bracket 402. The rotating platform 403 can drive the camera platform 404 to rotate. The signals transmitted by the controller unit 109 can control the rotating platform 403 to adjust the angle of the camera 401; at the same time, it can control the brightness of the auxiliary ring light 405 to ensure the clarity of the image information acquired by the camera 401. The acoustic sensor 410 is located above the housing connecting plate 407. During the operation of the pipeline robot, the acoustic sensor will emit sound wave signals according to the given frequency of the controller unit 109. When the sound wave signals encounter obstacles, they will return a response signal. The acoustic sensor collects the reflected sound waves in real time and converts the acoustic signals into electrical signals, which are then transmitted back to the controller unit 109. The controller unit 109 will obtain the distance between the robot and the pipeline endpoint based on the sound wave transmission time, assisting the gyroscope module in precise positioning. During the operation of the pipeline robot, the visual information acquisition unit 106 provides the pipeline robot with the function of acquiring image information.

[0043] Figure 5 The overall communication architecture of the pipeline inspection robot of this invention is shown in the figure. During the communication process, the ROS system running on the Raspberry Pi 5 development board distributes control signals to each module and receives data.

[0044] A vision-based inspection robot method for inspecting small-diameter pipes includes the following steps: Step 1: Place the pipeline robot at the entrance and complete the precise docking of the power supply module with the main equipment using a wired connection. Then, check the functional status of the main components such as the tracked drive unit, the lifting gimbal module, the vision information acquisition unit, and the controller unit one by one.

[0045] Step 2: The user uses the microphone embedded in the remote terminal device to issue natural language commands. The wired communication transmits the sound signal to the Raspberry Pi 5 development board. The speech recognition software on the development board converts the collected sound information into text and passes it to the Qwen-4B language model. After receiving the text input, the model completes semantic parsing based on the historical data and context information of the pipeline detection scenario, generates a complete command sequence, and converts the sequence into a machine execution format. Finally, the ROS system realizes the automated operation process.

[0046] Step 3: After receiving the instruction from the ROS system, the tracked drive unit drives the robot into the pipeline. The gyroscope sensor continuously sends the current angular velocity information to the control system. The searchlight provides basic illumination to the forward-looking area. The image acquisition unit acquires pipe wall image data according to a predetermined cycle and uploads it to the controller. The lifting platform changes the camera lens pitch angle according to the detection requirements.

[0047] Step 4: After receiving the image information captured by the visual information acquisition unit, the controller waits for the pipeline robot to stabilize before activating other sensor modules. The laser sensor scans the three-dimensional contour of the inner wall of the pipeline in real time, and the ultrasonic module works with the sound wave module to convert the vibration signal into an electrical signal and transmit it to the controller.

[0048] Step 5: The controller integrates the image features, laser size parameters, ultrasonic depth information and acoustic vibration data collected by each module, and then sends these data to the remote terminal display interface. It can also perform dynamic storage.

[0049] Step 6: The detection process ends. The user issues a "stop detection" voice command via microphone. Upon receiving the command, the Qwen-4B large model sends a "task abort" signal to the ROS system, causing the tracked drive unit to stop operating, the robot to enter a stationary state, the lifting platform to descend to its lowest point, the camera to return to its initial posture via the rotation servo motor, the laser, acoustic, and ultrasonic sensors to automatically turn off, and the searchlight equipment to turn off as well. The controller unit saves the collected data to the Raspberry Pi 5's SD card for later analysis and processing.

Claims

1. A vision inspection robot system for inspecting small-diameter pipes, characterized in that, It includes the following structure: main base, tracked drive unit, liftable gimbal, visual information acquisition unit, and controller unit. Tracked drive units are symmetrically mounted on both sides of the main body base; A liftable gimbal is provided at the center of the main base; The top of the liftable gimbal mechanism is equipped with a visual information acquisition unit; The liftable gimbal mechanism has a controller unit on its back.

2. The vision inspection robot system for small-diameter pipe inspection according to claim 1, characterized in that: The tracked drive unit includes bearings, a motor, tracks, and drive wheels. The drive wheels are located on both sides of the main body base, and there are two drive wheels on each side. The motor is mounted on the frame, and the output shaft of the motor is coupled to the bearings. The tracked drive unit is used to drive the robot to move inside the pipeline.

3. The vision inspection robot system for small-diameter pipe inspection according to claim 1, characterized in that: The height-adjustable gimbal module includes a rotating base, a fixed housing, and connecting rods. The fixed housing is rigidly connected to the upper surface of the rotating base. The side of the fixed housing is provided with several coaxial mounting holes adapted to the connecting rod structure. The connecting rod includes two sets of parallelogram components. Each set of components includes two long connecting rods of equal length. One end of the connecting rod is hinged to a preset hole on the side wall of the fixed housing via a pin, and the other end supports the visual information acquisition unit. The height-adjustable gimbal module provides an adjustable platform for the visual information acquisition unit.

4. The vision inspection robot system for small-diameter pipe inspection according to claim 1, characterized in that: The visual information acquisition unit includes a camera, a camera mounting bracket, a rotating platform, a camera platform, an auxiliary ring light, a lighting platform, a housing connecting plate, a supporting housing, and hinge pins. The hinge pins are installed on both sides of the supporting housing and are hinged to the liftable gimbal module. The housing connecting plate is fixedly connected to the supporting housing and also to the lighting platform. The auxiliary ring light is installed on the lighting platform, the camera platform is installed on the lighting platform, and the rotating platform is installed on the camera platform. The camera is fastened to the camera platform via the camera mounting bracket and is connected to the camera mounting bracket. The visual information acquisition unit acquires image data for the pipeline robot and sends it to the controller unit.

5. A vision inspection robot system for small-diameter pipe inspection according to claim 1, characterized in that: The controller unit uses a Raspberry Pi 5 development board, which locally deploys the Qwen-4B large language model and ROS system. The controller unit is electrically connected to the tracked drive unit, the liftable gimbal, and the visual information acquisition unit. The controller unit is used to send control signals to the other modules and receive data information from other modules.

6. A vision-based inspection robot method for inspecting small-diameter pipes, characterized in that, Includes the following steps: Step 1: Place the pipeline robot at the pipeline inlet and connect the power supply module to the main equipment via wired connection. Then, check the functional status of the main components such as the tracked drive unit, the liftable gimbal module, the visual information acquisition unit, and the controller unit one by one. Step 2: The user uses the microphone embedded in the remote terminal device to issue natural language commands. The wired communication transmits the sound signal to the Raspberry Pi 5 development board. The speech recognition software on the development board converts the collected sound information into text and passes it to the Qwen-4B language model. After receiving the text input, the model completes semantic parsing based on the historical data and context information of the pipeline detection scenario, generates a complete command sequence, and converts the sequence into a machine execution format. Finally, the ROS system realizes the automated operation process. Step 3: After receiving the instructions from the ROS system, the tracked drive unit drives the robot into the pipeline. The gyroscope sensor continuously sends the current angular velocity information to the control system. The searchlight provides basic illumination to the forward-looking area. The image acquisition unit acquires pipe wall image data according to a predetermined cycle and uploads it to the controller. The liftable gimbal can change the tilt angle of the camera lens according to the detection requirements. Step 4: After receiving the image information captured by the visual information acquisition unit, the controller waits for the pipeline robot to stabilize before activating other sensor modules. The laser sensor scans the three-dimensional contour of the inner wall of the pipeline in real time, and the ultrasonic module works with the sound wave module to convert the vibration signal into an electrical signal and transmit it to the controller. Step 5: The controller integrates the image features, laser size parameters, ultrasonic depth information and acoustic vibration data collected by each module, and then sends these data to the remote terminal display interface. It can also perform dynamic storage. Step 6: The detection process ends. The user issues a "stop detection" voice command via microphone. Upon receiving the command, the Qwen-4B large model sends a "task abort" signal to the ROS system, causing the tracked drive unit to stop operating, the robot to enter a stationary state, the lifting gimbal to descend to its lowest point, the camera to return to its initial posture via the rotating servo motor, the laser, acoustic, and ultrasonic sensors to automatically turn off, and the searchlight equipment to turn off as well. The controller unit saves the collected data to the Raspberry Pi 5's SD card for later analysis and processing.

Images