Intelligent pipeline detection robot integrating binocular vision and IMU (Inertial Measurement Unit) and implementation method of intelligent pipeline detection robot

Abstract

The invention relates to the technical field of intelligent pipeline detection robots, and discloses a binocular vision and IMU integrated intelligent pipeline detection robot and an implementation method thereof.The binocular vision and IMU integrated intelligent pipeline detection robot comprises a self-adaptive binocular vision module, an IMU control module, an IMU control module, an IMU control module and an IMU control module, and the environment perception model based on deep learning can automatically adjust camera parameters according to illumination conditions and surface reflection characteristics in the pipeline so as to obtain the optimal image quality. According to the intelligent pipeline detection robot fusing binocular vision and IMU and the implementation method thereof, through an environment perception model based on a convolutional neural network, the robot can analyze the illumination intensity, color distribution and surface texture in a pipeline in real time and automatically adjust the aperture and focal length of a camera, so that high-quality images are obtained under various illumination conditions, and the detection efficiency is improved. And the image quality is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent pipeline detection robots, and in particular to an intelligent pipeline detection robot integrating binocular vision and IMU and an implementation method thereof. Background Art

[0002] An intelligent pipeline inspection robot is an automated device designed specifically for inspecting and assessing the internal condition of pipelines. Incorporating a variety of advanced sensors and technologies, it enables efficient and accurate inspections in complex and narrow pipeline environments. However, existing binocular vision systems may perform poorly in low-light, highly reflective, or texture-poor environments, resulting in an inability to accurately obtain structural information about the pipeline's interior. Furthermore, in complex pipeline environments, existing robots may encounter various unforeseen situations, such as sudden obstacles or sensor failures, which existing algorithms are unable to cope with well. Summary of the Invention

[0003] The purpose of this section is to summarize some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the abstract and title of this application to avoid obscuring the purpose of this section, the abstract and the title of the invention, and such simplifications or omissions should not be used to limit the scope of the present invention.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] The intelligent pipeline inspection robot integrates binocular vision and IMU. Its modules include:

[0006] Adaptive binocular vision module: This module includes at least two cameras with adjustable aperture and focal length, and a deep learning-based environmental perception model that automatically adjusts camera parameters based on the lighting conditions and surface reflectivity inside the pipeline to achieve optimal image quality.

[0007] High-precision inertial measurement unit: used to provide high-precision attitude information, including a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer;

[0008] Data fusion processing unit: uses a multimodal data fusion algorithm based on deep learning to efficiently fuse the image information obtained by the binocular vision module with the posture information provided by the IMU, reducing data processing delays and improving positioning accuracy;

[0009] Autonomous navigation and obstacle avoidance module: This module includes a 3D map construction function based on the SLAM algorithm, as well as a real-time path planning and obstacle recognition algorithm based on deep learning, enabling autonomous navigation and intelligent obstacle avoidance.

[0010] Low power design: including an energy management system that extends the robot's operating time by dynamically adjusting the working status and power consumption of each component;

[0011] Remote monitoring and fault diagnosis module: supports multiple wireless communication protocols, allowing users to remotely monitor the robot's working status and receive fault diagnosis reports.

[0012] As a further solution of the present invention: the adaptive binocular vision module further includes:

[0013] Environmental perception model: This model, based on a convolutional neural network, analyzes the light intensity, color distribution, and surface texture inside the pipeline in real time, automatically adjusting the camera's aperture and focal length.

[0014] Image enhancement algorithms: used to improve image quality in low-light conditions, including noise reduction and contrast enhancement.

[0015] Furthermore, in the illumination intensity estimation, it is assumed that the illumination intensity I can be approximated by the image brightness B. The convolutional layer is used to extract the image feature F, and then the illumination intensity prediction value I^ is obtained through the fully connected layer W and the bias b;

[0016] formula:

[0017] Aperture and focus adjustment: adjust the aperture A and focal length F according to the light intensity I^ and the predefined mapping functions aperture(I) and ffocus(I);

[0018] formula:

[0019]

[0020] As a further solution of the present invention: the data fusion processing unit further includes:

[0021] IMU data correction algorithm: A Kalman filter-based algorithm is used to correct the drift of IMU data in real time to ensure the accuracy of attitude information during long-term operation;

[0022] Multimodal data fusion algorithm: A fusion algorithm based on deep learning that can simultaneously process image data and IMU data to improve the accuracy and efficiency of data fusion.

[0023] Furthermore, in the state equation, let xk be the state vector at the kth moment, uk be the control input, wk be the process noise, A be the state transfer matrix, and B be the control input matrix;

[0024] Formula: x k =Ax k-1 +Bu k +wk ;

[0025] In the observation equation, let zk be the observation vector at the kth moment, H be the observation matrix, and vk be the observation noise;

[0026] Formula: z k =Hx k +v k ;

[0027] Kalman gain: Let Kk be the Kalman gain, Pk be the error covariance matrix, R be the observation noise covariance matrix, and Q be the process noise covariance matrix;

[0028] Formula: K k =P k H T (HP k H T +R) -1 ;

[0029] Status update: x k =x k +K k (z k -Hx k );

[0030] P k =(IK k H)P k ;

[0031] The multimodal data fusion algorithm based on deep learning includes feature extraction. Let Fv be the image feature and Fi be the IMU feature. Use the convolution layer and the fully connected layer to extract Fv and Fi respectively. The formula is as follows:

[0032] F v =CNN v (I);

[0033] F i =FC i (IMU);

[0034] Feature fusion: Fuse the image features and IMU features to obtain the comprehensive feature Fc. Formula: F c =σ(W v F v +W i F i +b);

[0035] Among them, σ is the activation function, Wv and Wi are the weight matrices of image features and IMU features respectively, and b is the bias.

[0036] As a further solution of the present invention: the autonomous navigation and obstacle avoidance module further includes:

[0037] 3D map construction algorithm: The 3D map construction function based on the SLAM algorithm can generate high-precision 3D maps of the pipeline interior in real time;

[0038] Path planning algorithm: A real-time path planning algorithm based on deep learning that can dynamically plan the optimal path based on a three-dimensional map;

[0039] Obstacle recognition algorithm: Based on convolutional neural networks, the obstacle recognition algorithm can identify obstacles of different types and shapes and take corresponding obstacle avoidance measures;

[0040] Furthermore, the construction of a 3D map based on the SLAM algorithm includes pose estimation: Let T k is the robot pose at the kth moment, z k is the observed data, x k For the map point set, the extended Kalman filter is used for pose estimation;

[0041] ^

[0042] Formula: T k =EKF(T k-1 ,z k );

[0043] Map point update: Update the map point set xk according to the new pose T^k;

[0044] ^

[0045] Formula: x k =T k x k-1 +Δx k ;

[0046] Among them, Δxk is the newly observed map point.

[0047] As a further solution of the present invention: the low power consumption design further includes:

[0048] Dynamic power management: Dynamically adjusts the operating status of each component, such as the camera frame rate and processor frequency, based on task requirements and environmental conditions;

[0049] Energy recovery system: Utilizes the kinetic energy generated by the robot during movement to convert it into electrical energy, replenishing battery power and extending operating time.

[0050] The implementation method of the intelligent pipeline inspection robot integrating binocular vision and IMU is as follows:

[0051] Environmental perception and image acquisition: The adaptive binocular vision module automatically adjusts camera parameters based on the lighting conditions and surface reflectivity inside the pipeline to obtain high-quality images.

[0052] Posture information acquisition: Use high-precision IMU to provide robot posture information, including three-axis acceleration, three-axis angular velocity, and three-axis magnetic field strength;

[0053] Data fusion processing: The data fusion processing unit efficiently fuses image information with posture information, reducing data processing delays and improving positioning accuracy;

[0054] 3D map construction: Use SLAM algorithm to build a high-precision 3D map of the pipeline interior in real time;

[0055] Path planning and obstacle avoidance: Path planning algorithms and obstacle recognition algorithms based on deep learning enable autonomous navigation and intelligent obstacle avoidance;

[0056] Low power consumption management: Dynamically adjust the working status of each component through the energy management system to extend the robot's operating time;

[0057] Remote monitoring and fault diagnosis: The robot's working status and fault diagnosis reports are sent to users through the remote communication module.

[0058] As a further solution of the present invention: the environment perception and image acquisition steps further include:

[0059] Environmental perception model: A convolutional neural network-based model analyzes the light intensity, color distribution, and surface texture inside the pipeline in real time, automatically adjusting the camera's aperture and focal length.

[0060] Image enhancement algorithms: Improve image quality in low-light conditions, including noise reduction and contrast enhancement.

[0061] As a further solution of the present invention: the data fusion processing step further includes:

[0062] IMU data correction algorithm: Based on the Kalman filter algorithm, it corrects the drift of IMU data in real time to ensure the accuracy of attitude information during long-term operation;

[0063] Multimodal data fusion algorithm: A fusion algorithm based on deep learning that processes image data and IMU data simultaneously to improve the accuracy and efficiency of data fusion.

[0064] As a further solution of the present invention: the three-dimensional map construction algorithm: based on the three-dimensional map construction function of the SLAM algorithm, generates a high-precision three-dimensional map of the interior of the pipeline in real time;

[0065] Path planning algorithm: A real-time path planning algorithm based on deep learning that dynamically plans the optimal path;

[0066] Obstacle recognition algorithm: Based on convolutional neural network, the obstacle recognition algorithm can identify obstacles of different types and shapes and take corresponding obstacle avoidance measures;

[0067] Furthermore, the path planning algorithm includes a path cost function: let C(p) be the cost of path p, L(p) be the path length, and O(p) be the density of obstacles on the path;

[0068] Formula: C(p) = αL(p) + βO(p);

[0069] Among them, α and β are weight coefficients;

[0070] Path optimization: Use deep reinforcement learning algorithm to optimize the path. Formula:

[0071] As a further solution of the present invention: the remote monitoring and fault diagnosis step further includes:

[0072] Real-time data transmission: supports multiple wireless communication protocols to ensure stable data transmission in different environments;

[0073] Fault diagnosis algorithm: A machine learning-based fault diagnosis algorithm automatically detects and reports robot faults and provides maintenance recommendations.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] Using a convolutional neural network-based environmental perception model, the robot can analyze the light intensity, color distribution, and surface texture inside the pipeline in real time, automatically adjusting the camera's aperture and focal length to obtain high-quality images under various lighting conditions. This not only improves image quality but also enhances the robot's environmental adaptability, allowing it to operate normally even in low-light or overly bright environments. The 3D map-building function based on the SLAM algorithm can generate a high-precision 3D map of the pipeline interior in real time, providing the robot with a global view and enabling it to better understand its surroundings. The obstacle recognition algorithm based on the convolutional neural network can identify obstacles of different types and shapes and take appropriate obstacle avoidance measures. This significantly enhances the robot's autonomous navigation capabilities and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 This is a module connection diagram of an intelligent pipeline inspection robot that integrates binocular vision and IMU and its implementation method. DETAILED DESCRIPTION

[0077] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0078] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0079] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0080] Example

[0081] See also Figure 1 , is an embodiment of the present invention, which provides an intelligent pipeline inspection robot integrating binocular vision and IMU and an implementation method thereof, including:

[0082] Step 1: Preparation stage:

[0083] Camera: Sony IMX224 CMX sensor with adjustable aperture and focus;

[0084] I MU: Xsens MT i-10 series, including a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer;

[0085] Robot platform: equipped with necessary computing units and power management modules;

[0086] Software Development:

[0087] Operating system: ROS2Foxy F itzroy;

[0088] Programming language: Python 3;

[0089] Step 2: Environmental Perception and Image Acquisition:

[0090] Experimental setup:

[0091] Deploy the robot in a simulated pipeline environment with varying lighting conditions and surface reflectance characteristics;

[0092] Experimental steps:

[0093] Start the adaptive binocular vision module:

[0094] Light intensity estimation:

[0095] Input image size: 640x480 pixels;

[0096] Use pre-trained CNN model to extract image features;

[0097] Light intensity estimation results:

[0098] Assume W is a 128x128 weight matrix, b is a 128-dimensional bias vector, and F is a 128-dimensional feature vector;

[0099] Results: Light intensity I^≈150I ux;

[0100] Aperture and focus adjustment:

[0101] Adjust the aperture and focal length according to the light intensity I^ and a predefined mapping function;

[0102]

[0103] Result: Aperture A = 2.8, focal length F = 10mm;

[0104] Image enhancement algorithms:

[0105] For images under low-light conditions, image enhancement algorithms are used to improve image quality;

[0106] Result: Image brightness increased by 20% and contrast increased by 15%;

[0107] Step 3: Acquisition of posture information:

[0108] Experimental setup:

[0109] Record the posture changes of the robot during movement;

[0110] Experimental steps:

[0111] Start IMU and record data:

[0112] I MU data sampling frequency: 100 Hz;

[0113] data:

[0114] Acceleration: a = [0.2, -0.3, 9.8] m / s 2 ;

[0115] Angular velocity: ω = [-0.1, 0.05, 0.02] rad / s;

[0116] Magnetic field strength: m = [0.25, 0.1, -0.15] μT;

[0117] IMU data correction algorithm:

[0118] Use the Extended Kalman Filter (EKF) to correct IMU data;

[0119] Equation of state: x k =Ax k-1 +Bu k +w k ;

[0120] Among them, A is the state transfer matrix, B is the control input matrix, w k is the process noise;

[0121] Observation equation: z k =Hx k +v k ;

[0122] Where H is the observation matrix, v k is the observation noise;

[0123] Kalman gain: K k =P k H T (HP k H T +R) -1 , where P k is the error covariance matrix, R is the observation noise covariance matrix;

[0124] Status update: x k =x k +K k (z k -Hx k );

[0125] P k =(IK k H)P k ;

[0126] Result: Corrected posture information:

[0127] Acceleration: acorr = [0.22, -0.35, 9.82] m / s 2 ;

[0128] Angular velocity: ωcorr = [-0.12, 0.04, 0.01] rad / s;

[0129] Magnetic field strength: mcorr = [0.26, 0.12, -0.14] μT;

[0130] Step 4: Data fusion processing:

[0131] Experimental setup:

[0132] In the same simulation pipeline environment, the binocular vision module and IMU are enabled at the same time;

[0133] Experimental steps:

[0134] Fusion of image information and posture information:

[0135] Image feature extraction: F v =CNN v (I), where F v is the image feature vector;

[0136] IMU feature extraction: F i =FC i (IMU), where F i is the IMU feature vector;

[0137] Feature fusion: F c =σ(W v F v +W i F i +b), where σ is the activation function, W v W v and W i are the weight matrices of image features and IMU features respectively, and b is the bias vector;

[0138] Results: The fused feature vector Fc contains 256 elements, representing the fused multimodal data;

[0139] Step 5: 3D map construction:

[0140] Experimental setup:

[0141] Record all images and posture information obtained during the robot's movement;

[0142] Experimental steps:

[0143] SLAM algorithm implementation:

[0144] Use ORB-SLAM2 to implement SLAM algorithm;

[0145] Pose estimation: Among them, T k is the robot pose at the kth moment, z k is the observation data;

[0146] Map point update: Where Δx k is the newly observed map point;

[0147] Results: The generated 3D map contains 1000 map points and covers a 10-meter-long pipeline segment;

[0148] Step 6: Path planning and obstacle avoidance:

[0149] Experimental setup:

[0150] Deploy the robot in a simulated pipeline environment containing multiple obstacles;

[0151] Experimental steps:

[0152] Path planning algorithm implementation:

[0153] Use a DRL-based path planning algorithm;

[0154] Path cost function: C(p) = αL(p) + βO(p), where L(p) is the path length, O(p) is the obstacle density on the path, and α and β are weight coefficients;

[0155] Path optimization: Among them, p* is the optimal path;

[0156] Result: The optimal path planned is 15 meters long, passes through 3 turning points, and avoids 5 obstacles;

[0157] Obstacle recognition algorithm implementation: Use CNN-based obstacle recognition algorithm;

[0158] Example results: 5 different types of obstacles are identified, including bumps, depressions, and cracks;

[0159] Step 7: Low power management:

[0160] Experimental setup:

[0161] Test the robot's power consumption under different load conditions;

[0162] Experimental steps:

[0163] Dynamic power management: adjusts camera frame rate and processor frequency based on task requirements and environmental conditions;

[0164] Results: Under low-load tasks, the camera frame rate is 15fps and the processor frequency is 1GHz; under high-load tasks, the camera frame rate is 30fps and the processor frequency is 1.5GHz;

[0165] Energy recovery system: converts the kinetic energy of the robot into electrical energy during movement;

[0166] Result: During the 1-hour mission, the energy recovery system increased the battery charge by 10%;

[0167] Step 8: Remote monitoring and fault diagnosis:

[0168] Experimental setup: The robot is remotely monitored via wireless communication protocols;

[0169] Experimental steps:

[0170] Real-time data transmission: supports Wi-Fi, Bluetooth, 4G / 5G and other wireless communication protocols;

[0171] Results: In a 4G network environment, the average delay is less than 50ms;

[0172] Fault diagnosis algorithm implementation: Use SVM-based fault diagnosis algorithm;

[0173] Results: Three fault conditions were successfully detected and reported, including camera failure and IMU data anomaly.

[0174] Step 9: Experimental data analysis:

[0175] Improved image quality: Under different lighting conditions, the adaptive binocular vision module adjusts the aperture and focal length, improving image brightness by an average of 15% and contrast by 10%.

[0176] Attitude information accuracy: After correcting the IMU data using the Kalman filter, the attitude information error was reduced by 20%;

[0177] Improved positioning accuracy: After the data fusion processing unit fuses image information with IMU information, positioning accuracy is improved by 15%;

[0178] Autonomous navigation capability: In a complex environment with multiple obstacles, the robot successfully avoided 95% of them and completed navigation tasks along the intended path.

[0179] Power management effect: Under low-load tasks, power consumption is reduced by 30%. Under high-load tasks, power consumption increases by 10%, but some of the power loss is compensated by the energy recovery system.

[0180] Remote monitoring stability: A stable remote monitoring function is achieved, with an average delay of less than 50ms and a fault diagnosis accuracy rate of 90%.

[0181] In summary, through the above specific experimental data and analysis, the intelligent pipeline inspection robot integrating binocular vision and IMU has demonstrated significant creative and beneficial effects in many aspects. The effects not only enhance the robot's environmental adaptability, posture information accuracy, data processing efficiency, autonomous navigation capability, power consumption management, and remote monitoring performance, but also significantly improve the robot's reliability and work efficiency in practical applications. These designs and methods make this robot an advanced solution in the field of pipeline inspection.

[0182] It is important to note that the construction and arrangement of the present application shown in a number of different exemplary embodiments are merely illustrative. Although only a few embodiments are described in detail in this disclosure, it should be readily understood by those who refer to this disclosure that many modifications are possible (e.g., the size, scale, structure, shape and proportion of various elements, as well as parameter values ​​(e.g., temperature, pressure, etc.), mounting arrangements, use of materials, color, directional changes, etc.) without departing substantially from the novel teachings and advantages of the subject matter described in this application. For example, an element shown as integrally formed may be composed of multiple parts or elements, the position of the element may be inverted or otherwise changed, and the nature or number or position of the discrete elements may be altered or changed. Therefore, all such modifications are intended to be included within the scope of the present invention. The order or sequence of any process or method steps may be changed or reordered according to alternative embodiments. In the claims, any "means plus function" clause is intended to cover the structure described herein that performs the function, and is not only structurally equivalent but also equivalent structures. Other replacements, modifications, changes, and omissions may be made in the design, operating conditions, and arrangement of the exemplary embodiments without departing from the scope of the present invention. Therefore, the invention is not limited to the specific embodiments, but extends to various modifications that still fall within the scope of the appended claims.

[0183] Additionally, in order to provide a concise description of exemplary embodiments, all features of an actual embodiment (i.e., those features that are not relevant to the best mode presently contemplated for carrying out the invention or those that are not relevant to implementing the invention) may not be described.

[0184] It will be appreciated that in the development of any actual embodiment, as in any engineering or design project, numerous implementation-specific decisions may be made. Such a development effort may be complex and time-consuming, but will, for those of ordinary skill having the benefit of this disclosure, be a routine undertaking of design, fabrication, and production without undue experimentation.

[0185] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Intelligent pipeline inspection robot integrating binocular vision and IMU, characterized by: Its modules include: Adaptive binocular vision module: includes at least two cameras with adjustable aperture and focal length, and a deep learning-based environmental perception model that can automatically adjust camera parameters according to the lighting conditions and surface reflection characteristics inside the pipeline to obtain the best image quality; High-precision inertial measurement unit: used to provide high-precision attitude information, including three-axis accelerometer, three-axis gyroscope and three-axis magnetometer; Data fusion processing unit: uses a multimodal data fusion algorithm based on deep learning to efficiently fuse the image information obtained by the binocular vision module with the posture information provided by the IMU, reducing data processing delays and improving positioning accuracy; Autonomous navigation and obstacle avoidance module: including 3D map construction function based on SLAM algorithm, as well as real-time path planning and obstacle recognition algorithm based on deep learning, to achieve autonomous navigation and intelligent obstacle avoidance; Low power consumption design: including energy management system, which can extend the robot's operation time by dynamically adjusting the working status and power consumption of each component; Remote monitoring and fault diagnosis module: supports multiple wireless communication protocols, allowing users to remotely monitor the robot's working status and receive fault diagnosis reports.

2. The intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 1 is characterized in that: The adaptive binocular vision module also includes: Environmental perception model: A model based on a convolutional neural network that can analyze the light intensity, color distribution, and surface texture inside the pipeline in real time and automatically adjust the aperture and focal length of the camera; Image enhancement algorithms: used to improve image quality in low-light conditions, including noise reduction and contrast enhancement.

3. The intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 1 is characterized in that: The data fusion processing unit also includes: IMU data correction algorithm: Kalman filter-based algorithm used to correct the drift of IMU data in real time to ensure the accuracy of attitude information during long-term operation; Multimodal data fusion algorithm: A fusion algorithm based on deep learning that can process image data and IMU data simultaneously to improve the accuracy and efficiency of data fusion.

4. The intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 1 is characterized in that: The autonomous navigation and obstacle avoidance module also includes: 3D map construction algorithm: The 3D map construction function based on the SLAM algorithm can generate a high-precision 3D map of the interior of the pipeline in real time; Path planning algorithm: A real-time path planning algorithm based on deep learning that can dynamically plan the optimal path based on a three-dimensional map; Obstacle recognition algorithm: The obstacle recognition algorithm based on convolutional neural network can identify obstacles of different types and shapes and take corresponding obstacle avoidance measures.

5. The intelligent pipeline inspection machine integrating binocular vision and IMU according to claim 1 is characterized in that: The low power consumption design also includes: Dynamic power management: Dynamically adjust the working status of each component, such as the frame rate of the camera and the frequency of the processor, according to task requirements and environmental conditions; Energy recovery system: The kinetic energy of the robot during movement is converted into electrical energy to replenish the battery and extend the operation time.

6. A method for implementing an intelligent pipeline inspection robot integrating binocular vision and IMU, according to the intelligent pipeline inspection robot integrating binocular vision and IMU according to any one of claims 1 to 5, characterized in that: The implementation steps are as follows: Environmental perception and image acquisition: The adaptive binocular vision module automatically adjusts the camera parameters according to the lighting conditions and surface reflection characteristics inside the pipeline to obtain high-quality images; Posture information acquisition: Use high-precision IMU to provide robot posture information, including three-axis acceleration, three-axis angular velocity and three-axis magnetic field strength; Data fusion processing: The image information and posture information are efficiently fused through the data fusion processing unit to reduce data processing delay and improve positioning accuracy; 3D map construction: Use SLAM algorithm to build a high-precision 3D map inside the pipeline in real time; Path planning and obstacle avoidance: Path planning algorithm and obstacle recognition algorithm based on deep learning to achieve autonomous navigation and intelligent obstacle avoidance; Low power consumption management: Dynamically adjust the working status of each component through the energy management system to extend the robot's operating time; Remote monitoring and fault diagnosis: The robot's working status and fault diagnosis reports are sent to users through the remote communication module.

7. The method for implementing the intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 6 is characterized in that: The environmental perception and image acquisition steps also include: Environmental perception model: A model based on a convolutional neural network that analyzes the light intensity, color distribution, and surface texture inside the pipeline in real time, and automatically adjusts the aperture and focal length of the camera; Image enhancement algorithms: Improve image quality in low-light conditions, including noise reduction and contrast enhancement.

8. The method for implementing the intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 6 is characterized in that: The data fusion processing step also includes: IMU data correction algorithm: Based on the Kalman filter algorithm, the drift of IMU data is corrected in real time to ensure the accuracy of attitude information during long-term operation; Multimodal data fusion algorithm: A fusion algorithm based on deep learning that processes image data and IMU data simultaneously to improve the accuracy and efficiency of data fusion.

9. The method for implementing the intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 6 is characterized in that: The three-dimensional map construction algorithm: based on the three-dimensional map construction function of the SLAM algorithm, generates a high-precision three-dimensional map inside the pipeline in real time; Path planning algorithm: A real-time path planning algorithm based on deep learning that dynamically plans the optimal path; Obstacle recognition algorithm: The obstacle recognition algorithm based on convolutional neural network can identify obstacles of different types and shapes and take corresponding obstacle avoidance measures.

10. The method for implementing the intelligent pipeline inspection robot integrating binocular vision and IMU according to claim 6, characterized in that: The remote monitoring and fault diagnosis step also includes: Real-time data transmission: Supports multiple wireless communication protocols to ensure stable data transmission in different environments; Fault diagnosis algorithm: A machine learning-based fault diagnosis algorithm automatically detects and reports robot faults and provides maintenance recommendations.

Images