Wall-climbing robot system aiming at steel pipe truss structure coating detection and application method

Abstract

The invention relates to a wall-climbing robot system for detecting a coating of a steel pipe truss structure and an application method, and the system comprises a self-adaptive magnetic adsorption mobile platform which can be adaptively and stably adsorbed on steel trusses with different pipe diameters to execute an inspection task; the robot platform is provided with an autonomous positioning navigation system composed of a laser radar, binocular vision and an airborne computer, and the position of the robot is positioned in the steel pipe truss without satellite signals through real-time mapping and a positioning algorithm of tight coupling of laser and vision. The front part of the robot is provided with an electrically foldable surrounding type array camera, so that full-surface coverage type shooting of a steel pipe component is realized, and real-time identification of coating apparent defects is executed in combination with an artificial intelligence disease identification algorithm in an airborne computer; and the lower part of the robot is provided with a telescopic elastic electromagnetic eddy current probe for detecting the thickness of the coating. The method has the advantages that the apparent diseases and the thicknesses of coatings of steel structures such as ceilings of high-speed rail stations and ceilings of large stadiums can be rapidly detected.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of steel structure detection and robot detection system, and particularly relates to a wall-climbing robot system for steel pipe truss structure coating detection and an application method. BACKGROUND

[0002] Large-span public buildings such as high-speed railway station buildings and large-scale stadiums generally adopt spatial steel pipe truss structures as the main roof bearing system. Such structures have the advantages of reasonable stress, beautiful shape and large span, but also face severe durability and safety maintenance challenges. The steel pipe truss is exposed to complex environments for a long time, and bears wind load, temperature stress, vibration, humid air and even chemical corrosion, and the surface rust-proof coating is the first and most important barrier to protect the internal steel from corrosion and ensure the safety of the structure within the design life. Once the coating appears cracks, blisters, peeling and other visual defects, or the thickness is insufficient due to construction or aging, local corrosion will be quickly caused, which not only weakens the cross section of the member, but also may endanger the safety of the overall structure due to stress concentration. Therefore, regular and comprehensive visual defect and thickness detection of the coating of such truss structures is an indispensable key link in facility operation and maintenance.

[0003] However, there are many technical difficulties in efficient detection of the coating of such special structures, and traditional methods face great challenges. First, the detection environment is extremely harsh: the roof truss is usually dozens of meters high, with a dense steel pipe grid structure inside, and the space is narrow and uneven, which is a typical "unstructured environment". Personnel high-altitude operation is extremely risky and inefficient; second, satellite positioning signals are invalid: in the closed or semi-closed station building and stadium, GPS, BDS and other satellite signals are severely blocked or completely invalid, making it impossible for automated equipment relying on such signals to perform autonomous positioning and navigation; third, the detection requirements are multiple and complex: complete coating detection not only needs to find visible visual defects, but also needs to measure the coating thickness quantitatively, which requires the detection platform to not only realize large-scale visual survey, but also to perform contactless detection on specific points.

[0004] To address these challenges, researchers at home and abroad have attempted to introduce robotic technology into this field. The current mainstream exploration direction focuses on unmanned aerial vehicle platforms. For example, Skygauge Robotics Company developed a multi-rotor unmanned aerial vehicle equipped with a thickness gauge, which can move parallel to the detection surface while maintaining its attitude through a clever tilting rotor design to achieve contact measurement. Many academic researches also focus on the control strategy of unmanned aerial vehicle contact detection, such as Gonzalez et al. calculating the relative pose of the unmanned aerial vehicle and the target through a distance sensor to plan the approach path, and Kocer, Fumagalli team is committed to solving the problem of accurate control and stability of unmanned aerial vehicle in physical contact state. In addition, to enhance the anti-collision ability in narrow space, Sanchez et al. designed an unmanned aerial vehicle model wrapped in a protective frame.

[0005] Although the above researches lay the foundation for air contact detection, their application paradigm is mainly aimed at relatively open and flat bridges or building facades. When directly transplanted to the interior of dense steel pipe trusses, there are significant limitations. First, the safety is poor. Unmanned aerial vehicles flying between dense members are prone to collision and crash, with high risk. Second, the stability is insufficient. Multi-rotor unmanned aerial vehicles need to consume energy continuously to resist probe pressure when contact detection is performed, and it is difficult to maintain absolute stability, which affects the thickness measurement accuracy, and wind disturbance has a huge impact on it. Positioning and navigation rely on external signals, which is also a problem. Most of the above-mentioned solutions assume that satellite signals or pre-set markers are available for positioning, and their autonomous operation capability is greatly reduced in complex truss internal environments without GPS.

[0006] Therefore, the industry urgently needs a new detection solution that can adapt to complex truss structures, break away from dependence on external signals, and have both apparent inspection and accurate thickness measurement capabilities. Wall-climbing robots, as a mobile platform with adhesion capability, are theoretically more suitable for long-term, close-range fine work on the surface of such structures. However, traditional wall-climbing robots are mostly designed for flat walls, and when applied to steel truss environments with dramatic changes in pipe diameter, raised welds, and complex nodes, they face a series of new challenges such as adhesion reliability, obstacle crossing ability, movement and turning on curved pipe surfaces, and accurate positioning in signal-free environments.

[0007] In summary, for the coating detection of steel pipe truss structures such as high-speed railway station buildings and large sports venues, existing technologies have obvious deficiencies in environmental adaptability, work safety, functional completeness, and positioning autonomy. SUMMARY

[0008] The wall-climbing robot system for coating detection of a steel pipe truss structure and the application method can realize integrated and automatic precise detection of the apparent defects and thickness conditions of the coating of the steel structure, effectively improve the detection efficiency and data manageability, and serve daily inspection and regular detection of large steel structure facilities.

[0009] The object of the present application is achieved by the following technical solutions: A wall-climbing robot system for coating detection of a steel pipe truss structure, characterized in that it comprises an adaptive magnetic adsorption moving platform, a multi-sensor tightly coupled positioning and mapping module, an electrically driven ring-shaped image acquisition module, an embedded defect identification module, a coating thickness detection module, and a data fusion and report generation module, wherein: The adaptive magnetic adsorption moving platform is used to adsorb the wall-climbing robot body on the surface of the steel pipe truss structure and realize movement of the wall-climbing robot body on the surface of the steel pipe truss structure. The multi-sensor tightly coupled positioning and mapping module is mounted on the adaptive magnetic adsorption moving platform, and determines the position during the robot inspection process in a satellite signal-free environment through a real-time positioning and mapping algorithm based on laser radar and binocular vision fusion. The electrically driven ring-shaped image acquisition module is mounted on the adaptive magnetic adsorption moving platform and is used to realize image acquisition of the surface of the steel pipe truss structure. The embedded defect identification module is mounted on the adaptive magnetic adsorption moving platform, performs real-time defect identification on the images acquired by the electrically driven ring-shaped image acquisition module, and detects the apparent defects of the coating. The coating thickness detection module is mounted on the adaptive magnetic adsorption moving platform and is used to perform thickness detection of the surface coating of the steel pipe truss structure. The data fusion and report generation module is connected to the embedded defect identification module and the coating thickness detection module, integrates the apparent defect data of the coating and the thickness data of the coating, and generates a detection report.

[0010] The adaptive magnetic adsorption moving platform comprises a main drive wheel with an adaptive inclination angle, a steering tail wheel, and a fuselage, wherein the main drive wheel and the steering tail wheel are both made of a material that can provide a magnetic adsorption and pulling force, and the main drive wheel and the steering tail wheel are respectively hinged to the fuselage, and the main drive wheel is connected to a driving device.

[0011] The multi-sensor tight coupling positioning and mapping module comprises a non-repeating spherical laser radar, an embedded inertial navigation module, a binocular camera and an on-board computer, and a real-time mapping and positioning algorithm based on multi-source data tight coupling is embedded in the on-board computer to construct a three-dimensional map of the surrounding of the steel pipe truss in real time using data of the three sensors of laser point cloud, image and inertial navigation, and to locate the position of the robot.

[0012] The electric folding ring image acquisition module comprises an electric folding lead screw and detection array cameras arranged at both ends and in the middle of the electric folding lead screw, and the electric folding lead screw changes the angle of the detection array cameras at both ends by being pushed by a lead screw motor at the hinge.

[0013] The coating thickness detection module comprises an electromagnetic induction probe based on the magnetic vortex principle, a pressure sensing unit and an electrically controlled telescopic platform, the electrically controlled telescopic platform connects the electromagnetic induction probe and drives the electromagnetic induction probe to contact the surface of the steel pipe truss structure, and the pressure sensing unit is used to control the contact force of the electromagnetic induction probe.

[0014] An application method of the wall climbing robot system for coating detection of a steel pipe truss structure, characterized in that the application method comprises the following steps: S1, placing the wall climbing robot body on the surface of the steel pipe truss structure, adsorbing on the surface of the steel pipe truss structure through the main drive wheel and the steering tail wheel of the self-adaptive magnetic adsorption moving platform, and simultaneously adjusting the inward inclination angle of the main drive wheel according to different pipe diameters, and flexibly adjusting the inspection direction through the omni-directional steering tail wheel at narrow parts such as nodes to perform inspection work; S2, using the laser radar, binocular camera and on-board computer carried by the wall climbing robot, determining the position of the robot during the inspection process in a satellite signal-free environment through a real-time positioning and mapping algorithm based on the fusion of laser radar and binocular vision; S3, during the inspection process of the wall climbing robot, controlling the electric folding ring array camera to adjust the shooting angle to realize full-surface coverage image acquisition of different pipe diameter steel pipe components; S4, using an artificial intelligence disease identification algorithm embedded in the on-board computer to perform real-time disease identification on the images collected by the array camera to detect coating surface defects; S5, during the inspection process of the robot, setting a detection interval time for the coating thickness, synchronously controlling the electromagnetic induction probe to adhere to the surface of the steel pipe truss structure, and performing coating thickness detection; S6, integrating the surface defect and thickness data to generate a detection report.

[0015] In step S2, the algorithm adopts a tightly coupled error state iterative Kalman filter to fuse lidar, camera and inertial navigation data, which takes the initial state as the global coordinate system, estimates the pose through state transition, and updates the state and covariance through forward propagation, and in the measurement update stage, the lidar point cloud is de-warped and frame-map matched, and the local plane is fitted by projecting the nearest neighbor points; the visual part performs sparse direct image alignment by minimizing photometric error; for map management, the lidar map adopts incremental KD-tree structure; the visual map adopts axis-aligned voxel hash table to efficiently manage map points and associate multi-view image blocks.

[0016] In step S4, the artificial intelligence disease identification algorithm includes a multi-scale feature fusion module, which aggregates feature maps from different levels of the backbone network and adopts an adaptive spatial feature fusion mechanism to effectively identify various coating defects with large scale differences; for the diversity of coating disease morphology, the network head adopts a double-branch detection structure, one branch performs class and position prediction based on anchor boxes, and the other branch directly regresses disease boundaries using an anchor-free mechanism; the outputs of the two branches are fused after non-maximum suppression to simultaneously and efficiently detect multiple diseases such as coating hollowing, peeling and recession.

[0017] The point cloud data obtained in step 2 is reconstructed into a high-precision three-dimensional model of the steel pipe truss structure, and the apparent defect categories and position information identified in step 4 and the coating thickness data measured in step 5 are mapped onto the three-dimensional model according to the positioning information obtained during robot collection, realizing accurate matching and fusion of detection data and three-dimensional spatial position; further, the system provides a visual human-computer interaction interface, supports layered and regional rendering and display of different types of diseases and thickness distribution in the three-dimensional model, and has defect labeling, thickness overrun warning, corrosion development trend analysis and report one-key export functions, thereby realizing comprehensive evaluation and digital management of the coating state.

[0018] During the movement of the wall climbing robot, it turns through the cooperation of the differential of the main drive wheels and the lateral rotation of the steering tail wheels.

[0019] The advantages of the present application are: 1) By integrating an adaptive magnetic adsorption mobile platform, multi-sensor positioning, variable angle image acquisition and automatic thickness detection function, real-time identification of apparent diseases can be realized, and accurate measurement of coating thickness can be completed simultaneously, truly realizing integrated comprehensive detection of the health status of steel structure coating.

[0020] 2) The tightly coupled positioning algorithm of lidar, vision and inertial navigation can construct a map and accurately position in real time in a satellite signal-free environment, solving the problem of autonomous navigation and pose determination in complex spatial structures.

[0021] 3) By designing an electric folding ring array camera, the shooting angle can be automatically adjusted according to the component geometry, realizing full-coverage image acquisition of different pipe diameter rods and planar components, and significantly improving the detection efficiency and data integrity.

[0022] 4) The electric control telescopic electromagnetic eddy current probe is integrated with a pressure feedback unit, which can automatically touch and adjust the contact force in real time during movement, protecting the coating and realizing high-throughput and non-destructive thickness measurement.

[0023] 5) Based on real-time positioning and three-dimensional point cloud reconstruction technology, all detection results are mapped into a unified digital model, supporting visual correlation analysis, historical data comparison and automatic report generation of diseases and thickness, greatly improving the usability and decision support capability of the detection results.

[0024] 6) The comprehensive detection can be completed by one robot and one crawling operation, which significantly improves the detection efficiency, reduces the risk of high-altitude operation and labor cost, and is especially suitable for rapid inspection and regular monitoring of large-span and complex steel structure facilities. BRIEF DESCRIPTION OF DRAWINGS

[0025] Figure 1 is a composition diagram of the wall climbing robot system of the present application; Figure 2 is a schematic diagram of the adaptive tilt angle drive wheel of the wall climbing robot system of the present application under different pipe diameters; Figure 3 is a schematic diagram of the array camera detecting different component surfaces of the present application; Figure 4 is a structural schematic diagram of the coating thickness detection module of the present application. DETAILED DESCRIPTION

[0026] The features and other related features of the present application are further described in detail below by embodiments combined with the drawings, so as to facilitate the understanding of the same by the skilled in the art: As shown in Figures 1-4 , the marks in the figure represent: main drive wheel 1, drive motor 2, steering tail wheel 3, non-repetitive spherical laser radar 4, on-board computer 5, telescopic lead screw 6, electric folding lead screw 7, detection array camera 8, binocular camera 9, electromagnetic induction probe 10, pressure sensing unit 11, robot mounting plate 12, telescopic rod 13.

[0027] Embodiment: As shown in Figures 1 to 4 , the application method of the wall climbing robot system for steel pipe truss structure coating detection in the present embodiment includes the following steps: Step 1: The wall climbing robot is adsorbed on the surface of the steel pipe truss, and the inspection is started.

[0028] In this embodiment, the wall-climbing robot platform adopts a moving mechanism with self-adaptive magnetic suction main drive wheels 1 and omnidirectional steering tail wheels 3. The main drive wheels 1 and the steering tail wheels 3 are made of neodymium-iron-boron permanent magnet material, which can provide a total magnetic suction force of not less than 120 kg, much larger than the weight of the robot (about 30 kg), ensuring reliable adsorption in the case of vertical wall surface and hanging. The main drive wheels 1 on both sides are driven by the drive motor 2 through belt transmission, and the whole power module is hinged to the body, which can automatically adjust the inward or outward angle according to the change of pipe diameter during inspection, ensuring that the magnetic wheels are in full contact with the surface of the steel structure at all times. In the narrow or steering area such as pipe nodes, flexible steering is realized by differential cooperation between the main drive wheels 1 on both sides and the transverse rotation of the steering tail wheel 3. The steering tail wheel 3 can provide transverse driving force to assist the robot to smoothly transition to the adjacent rod with a steering angle of less than 30°.

[0029] Step 2: Use multi-sensor fusion positioning algorithm to determine the robot position in real time and build an environment map.

[0030] The robot is equipped with a non-repetitive scanning laser radar 4, a binocular camera 9 and an inertial measurement unit. The on-board computer runs a real-time positioning and mapping algorithm based on a tightly coupled error state iterative Kalman filter. The algorithm takes the initial state as the world coordinate system, predicts the pose through state transfer and forward propagation; in the measurement update stage, the laser point cloud is deformed and frame-map matched, and the local plane is fitted by projecting the nearest neighbor points, while the vision part realizes direct image alignment by minimizing the photometric error. The laser map uses incremental KD-tree management, and the visual map uses axis-aligned voxel hash table to store multi-view image blocks. This system can achieve centimeter-level positioning accuracy in GPS-free environment, and is suitable for complex scenes such as light change and texture loss.

[0031] Step 3: Control the electric folding ring-encircling array camera to adjust the angle and realize full-coverage image acquisition.

[0032] The array camera module is composed of three detection array cameras 8 of the same specification, installed at both ends and the middle of the electric folding lead screw 7. As shown in Figure 3 When detecting planar members (such as steel plates), the three cameras are in the same horizontal plane to expand the horizontal field of view; when detecting circular or rectangular pipes, the angle of the monitoring array cameras 8 at both ends is adjusted by the screw motor of the electric folding lead screw 7 to make them in a ring-encircling layout on the circular or rectangular pipe, synchronously collecting images of the three side surfaces of the rod, realizing full-coverage shooting of the rod surface without repeated adjustment of the robot position, and significantly improving the detection efficiency. In this embodiment, the middle part of the electric folding lead screw 7 is connected with a telescopic lead screw 6, which is used to control the extension or retraction of the electric folding lead screw 7.

[0033] Step 4: Real-time identification of coating apparent diseases through embedded artificial intelligence algorithm.

[0034] The airborne computer 5 carried by the robot is a high-performance AI edge computing processor (computing power >=100 TOPS), which runs a lightweight target detection network. The network takes improved MobileNetV3 as the backbone, embeds the coordinate attention mechanism, and enhances the capture ability of small disease characteristics; the neck adopts a multi-scale feature fusion module, which aggregates different level features through an adaptive spatial feature fusion mechanism; the detection head adopts a double-branch structure, one branch predicts the disease class and position based on anchor box, and the other branch directly regresses the bounding box using anchor-free mechanism, the outputs of the two branches are fused through non-maximum suppression to realize high-precision real-time identification of multiple diseases such as rust, peeling, cracking and hollowing.

[0035] Step 5: Control the electrically controlled telescopic electromagnetic eddy current probe to perform coating thickness measurement.

[0036] The coating thickness detection module is installed on the bottom of the robot body through the robot mounting plate 12, and the electromagnetic induction probe 10 adopting the magnetic eddy current principle is installed on the electrically controlled linear telescopic platform with a telescopic rod 13. The electromagnetic induction probe 10 integrates a pressure sensing unit 11, which can feedback the contact force in real time during measurement, and maintain the pressure within the set threshold through closed-loop control to avoid pressure loss of the coating. The triggering frequency of the telescopic rod 13 is linked with the robot moving speed to realize sampling at fixed spatial intervals. The coating thickness data is transmitted to the main control system in real time, and is bound with the positioning information to generate a thickness spatial distribution map.

[0037] Step 6: Integrate multi-source data and generate a visual detection report based on the three-dimensional model.

[0038] Based on the point cloud data obtained in step 2, a high-precision three-dimensional model of the steel pipe truss is reconstructed, and the position, category information and thickness measurement data of the apparent disease are mapped into the model. The system provides a human-computer interaction interface, supports layered rendering and query according to disease type, thickness interval, member number and other dimensions, has thickness overrun warning, corrosion area trend analysis and historical data comparison functions, and can generate a structured detection report (supports PDF, Excel and other formats) one key, providing comprehensive data support for maintenance decision-making.

[0039] The wall climbing robot system in the embodiment of the application comprises: 1. An adaptive magnetic adsorption mobile platform with a tilt angle self-adjusting magnetic drive wheel and an omnidirectional steering tail wheel.

[0040] 2. A multi-sensor tightly coupled positioning and mapping module, including a laser radar, a vision, an IMU and an embedded positioning algorithm.

[0041] 3. An electrically driven ring-shaped image acquisition module with an adjustable array camera architecture.

[0042] 4. Embedded disease identification module, including AI processor and lightweight deep learning network.

[0043] 5. Coating thickness detection module, including magnetic eddy current probe, electric control telescopic platform and pressure feedback unit.

[0044] 6. Data fusion and report generation module, used for three-dimensional modeling, data fusion and visual output.

[0045] The embodiment realizes efficient, accurate and automatic detection of the coating condition of the steel pipe truss structure through the above system and method, and is particularly suitable for in-service inspection and regular detection of large and complex steel structure facilities.

[0046] Although the above embodiments have been described in detail with reference to the accompanying drawings for the purpose of conception and embodiment of the present application, those skilled in the art can recognize that various improvements and changes can be made to the present application without departing from the scope defined by the claims, and therefore, detailed description is not given here.

Claims

1. A wall-climbing robot system for coating inspection of a steel pipe truss structure, characterized by: The application relates to a wall-climbing robot, which comprises an adaptive magnetic adsorption moving platform, a multi-sensor close-coupling positioning and mapping module, an electric ring-type image acquisition module, an embedded disease identification module, a coating thickness detection module and a data fusion and report generation module. The adaptive magnetic adsorption moving platform is used for adsorbing a wall-climbing robot body on the surface of a steel pipe truss structure and realizing movement of the wall-climbing robot body on the surface of the steel pipe truss structure. The multi-sensor close-coupling positioning and mapping module is mounted on the adaptive magnetic adsorption moving platform, and determines the position of the robot during the inspection in a satellite signal-free environment through a real-time positioning and mapping algorithm based on laser radar and binocular vision fusion. The electric ring-type image acquisition module is mounted on the adaptive magnetic adsorption moving platform, and is used for realizing image acquisition of the surface of the steel pipe truss structure. The embedded disease identification module is mounted on the adaptive magnetic adsorption moving platform, and realizes real-time disease identification of the images acquired by the electric ring-type image acquisition module, and detects coating surface defects. The coating thickness detection module is mounted on the adaptive magnetic adsorption moving platform, and is used for performing thickness detection of the surface coating of the steel pipe truss structure. The data fusion and report generation module connects the embedded disease identification module and the coating thickness detection module, integrates coating surface defect data and coating thickness data to generate a detection report.

2. The wall-climbing robot system for coating inspection of a steel pipe truss structure according to claim 1, characterized in that: The adaptive magnetic adsorption moving platform comprises a main driving wheel with an adaptive inclination angle, a steering tail wheel and a machine body, wherein the main driving wheel and the steering tail wheel are both made of a material capable of providing magnetic adsorption and release, and are respectively hinged to the machine body, and the main driving wheel is connected to a driving device.

3. The wall-climbing robot system for coating inspection of steel pipe truss structure according to claim 1, characterized in that: The multi-sensor close-coupling positioning and mapping module comprises a non-repeated spherical laser radar, an embedded inertial navigation module and a computer on board, wherein a real-time mapping and positioning algorithm based on multi-source data close coupling is embedded in the computer on board, and the algorithm uses data of three sensors, i.e., laser point cloud, images and inertial navigation, to construct a three-dimensional map of the periphery of the steel pipe truss in real time, and to locate the position of the robot.

4. The wall-climbing robot system for coating inspection of steel pipe truss structure according to claim 1, characterized in that: The electric ring-type image acquisition module comprises an electric folding lead screw and detection array cameras arranged at both ends and in the middle of the electric folding lead screw, and the electric folding lead screw changes the angle of the detection array cameras at both ends through a lead screw motor at the hinge.

5. The wall-climbing robot system for coating inspection of steel pipe truss structure according to claim 1, characterized in that: The coating thickness detection module comprises an electromagnetic induction probe based on the magnetic vortex principle, a pressure sensing unit and an electrically-controlled telescopic platform, the electrically-controlled telescopic platform is connected to the electromagnetic induction probe and drives the electromagnetic induction probe to contact the surface of the steel pipe truss structure, and the pressure sensing unit is used for controlling the contact force of the electromagnetic induction probe.

6. A method of using the wall-climbing robot system for coating inspection of a steel pipe truss structure according to any one of claims 1 to 5, characterized by: The application method comprises the following steps: S1, placing a wall-climbing robot body on the surface of a steel pipe truss structure, adsorbing the wall-climbing robot body on the surface of the steel pipe truss structure through the main driving wheel and the steering tail wheel of the adaptive magnetic adsorption moving platform, and adaptively adjusting the inner inclination angle of the main driving wheel according to different pipe diameters, and flexibly adjusting the inspection direction through the omnidirectional steering tail wheel at narrow parts such as nodes, and performing inspection work. S2, the laser radar, binocular camera and on-board computer carried by the wall climbing robot are used to determine the position of the robot during the inspection process in the absence of satellite signals through the real-time positioning and mapping algorithm of laser radar and binocular vision fusion; S3, during the inspection process of the wall climbing robot, the shooting angle of the electric folding ring-encircling array camera is controlled to adjust, so that the image acquisition of the full surface of the steel pipe component with different diameters is realized; S4, the real-time disease identification of the image collected by the array camera is performed through the artificial intelligence disease identification algorithm embedded in the on-board computer to detect the surface defects of the coating; S5, during the inspection process of the robot, the detection interval time of the coating thickness is set, the surface of the steel pipe truss structure is synchronously controlled by the electromagnetic induction probe to press and adhere, and the coating thickness detection is performed; S6, the detection report is generated by integrating the surface defect and thickness data.

7. The method of claim 6, wherein the method is applied to a wall-climbing robot system for coating inspection of a steel pipe truss structure. In step S2, the algorithm uses a tightly coupled error state iterative Kalman filter to fuse laser radar, camera and inertial navigation data, which takes the initial state as the global coordinate system, estimates the pose through state transition, and updates the state and covariance with the help of forward propagation. In the measurement update stage, the laser radar point cloud is deformed and frame-map matched, and the local plane is fitted by projecting the nearest neighbor points; the vision part performs sparse direct image alignment by minimizing the photometric error; in terms of map management, the laser radar map adopts incremental KD tree structure; the visual map adopts axis-aligned voxel hash table to efficiently manage map points and associate multi-view image blocks.

8. The method of claim 6, wherein the method is applied to a wall-climbing robot system for coating inspection of a steel pipe truss structure. In step S4, the artificial intelligence disease identification algorithm includes a multi-scale feature fusion module, which aggregates feature maps from different levels of the backbone network and uses an adaptive spatial feature fusion mechanism to effectively identify various coating defects with large scale differences. For the diversity of coating disease morphology, the network head uses a double-branch detection structure, one branch based on anchor box for class and position prediction, and the other branch using anchor-free mechanism to directly regress disease boundary. The outputs of the two branches are fused after non-maximum suppression to efficiently detect multiple diseases such as coating hollowing, peeling and recession.

9. The method of claim 6, wherein the method is applied to a wall-climbing robot system for coating inspection of a steel pipe truss structure. The point cloud data obtained in step 2 is reconstructed into a high-precision three-dimensional model of the steel pipe truss structure, and the surface defect categories, position information identified in step 4 and the coating thickness data measured in step 5 are mapped to the three-dimensional model according to the positioning information during the robot collection, realizing accurate matching and fusion of detection data and three-dimensional space position; further, the system provides a visual human-computer interaction interface, supports layered and regional rendering display of different types of diseases and thickness distribution in the three-dimensional model, and has defect labeling, thickness overrun warning, corrosion development trend analysis and report one-key export functions, thereby realizing comprehensive evaluation and digital management of the coating state.

10. The method of claim 6, wherein the method is applied to a wall-climbing robot system for coating inspection of a steel pipe truss structure. During the movement of the wall climbing robot, it turns through the cooperation of the differential of the main drive wheels and the lateral rotation of the steering tail wheels.

Images