A visual detection method and device for a ship segment large-plane laser derusting wall-climbing robot
The laser-guided rust removal robot, which utilizes visual inspection and closed-loop control, enables precise identification and differentiated cleaning of rusted areas. This solves the problems of unstable rust removal quality and low efficiency in existing technologies, and improves the automation level of rust removal on large flat surfaces of ship sections.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHIPBUILDING TECHNOLOGY RESEARCH INSITITUTE (NO 11 INSTITUTE OF CSSC)
- Filing Date
- 2026-03-30
- Publication Date
- 2026-07-10
Smart Images
Figure CN122368942A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of shipbuilding technology, specifically to a visual inspection method and device for a large-scale laser rust removal climbing robot for ship sections. Background Technology
[0002] Sectional manufacturing is a core part of shipbuilding and repair, and the quality of secondary rust removal on the large surface areas of these sections directly affects the adhesion of subsequent coatings and the overall corrosion resistance of the ship. Traditional rust removal methods for section surfaces mainly employ sandblasting, high-pressure jetting, or manual grinding, which suffer from serious dust pollution, harsh working environments, high labor intensity, low efficiency, and difficulty in waste recycling, making it difficult to meet the environmental protection and automation requirements of modern green shipbuilding.
[0003] In recent years, laser cleaning technology has been gradually applied in the field of ship rust removal due to its advantages such as non-contact operation, no pollution, good controllability, and minimal damage to the substrate. Combining laser cleaning technology with a wall-climbing robot platform to form a laser rust removal wall-climbing robot has become an important technological direction for realizing automated rust removal of large flat surfaces in ship sections.
[0004] The existing technology still has the following shortcomings in practical applications: In terms of rust condition identification, existing laser rust removal climbing robots mostly operate using preset fixed process parameters, repeatedly scanning the same area multiple times in a single mode, resulting in low energy efficiency. The rust state of ship sections varies significantly during actual production, and operating with fixed parameters easily leads to under- or over-rust removal, affecting rust removal quality and substrate safety.
[0005] In terms of work area planning, existing technologies typically use full-coverage scanning. Wall-climbing robots often perform laser cleaning on areas without rust, resulting in unnecessary energy consumption and extended work time, which reduces the robot's work efficiency and energy utilization.
[0006] In terms of rust removal effectiveness inspection, existing processes and equipment are mostly open-loop controlled, lacking a real-time feedback mechanism for rust removal quality. After rust removal is completed, manual sampling or offline testing is required, which cannot promptly detect and correct rust removal defects, increasing rework rates and manufacturing costs.
[0007] In terms of the reliability of vision systems, the high-intensity reflected light generated during laser cleaning poses a serious threat to the lenses of industrial cameras. Long-term exposure to the laser operating environment will cause irreversible damage to the lens's photosensitive elements, affecting the quality of subsequent image acquisition.
[0008] Therefore, developing a control method for a laser rust removal climbing robot that can perceive rust conditions in real time, intelligently match process parameters, accurately plan the work area, inspect the rust removal effect online, and ensure the reliable operation of the vision system is of great significance for improving the intelligence level of rust removal operations on large flat surfaces of ship sections, reducing manufacturing costs, and promoting the large-scale application of green cleaning technology in the shipbuilding and repair field. Summary of the Invention
[0009] This invention proposes a visual inspection method and device for a large-scale laser rust removal climbing robot for ship sections. It overcomes the shortcomings of existing laser rust removal climbing robots, such as lack of adaptability to working conditions, rough planning of the working area, inability to provide real-time feedback on rust removal effect, and insufficient reliability of the vision system, and achieves efficient and energy-saving automated rust removal operations.
[0010] To achieve the above objectives, the technical solution of the present invention is as follows: A visual inspection method for a wall-climbing robot used for large-area laser rust removal in ship sections, characterized by the following steps: Step S1) Establish a database of typical planar corrosion images of ship sections; Original images of planar corrosion of ship sections were collected, and the corrosion areas and joint areas in the images were manually labeled to construct a labeled dataset containing different corrosion types and different lighting conditions. Step S2) After the wall-climbing robot reaches the work site, it uses the onboard industrial camera to collect real-time planar images of the ship sections in the current work area; Step S3) The industrial camera performs rust segmentation and working condition identification on the real-time acquired images. Through the pre-trained and fine-tuned image segmentation model SAM model, the rust area in the image is segmented at the pixel level, and the working condition identification of the working area is divided into four categories: closure seam rust, thick rust, light surface rust, and no rust removal required. Step S4) Based on the working condition category identified by the industrial camera, match the preset laser cleaning process parameter library to match the corresponding laser cleaning mode and process parameters for different working conditions; Step S5) Before starting the laser cleaning operation, perform the industrial camera protection action by adjusting the position of the industrial camera with the electric pan-tilt head so that the photosensitive element of the industrial camera faces away from the laser operation area. Step S6) Based on the pixel-level segmentation results of the rusted area, determine the ROI area for the rust removal operation, calculate the actual working stroke range of the laser head, plan the laser scanning trajectory, and execute the laser cleaning operation. Step S7) After the laser cleaning operation is completed, the industrial camera is reset to the shooting position by controlling the electric pan-tilt head, and the image of the cleaned work area is collected to inspect the rust removal effect online. The inspection results are fed back to the wall-climbing robot control system to form a closed-loop control.
[0011] Furthermore, in step S1, the labeled dataset includes positive rust samples and negative samples of paint / stains / artificial chalk markings.
[0012] Furthermore, in step S2, an industrial camera is installed at the front end of the wall-climbing robot. The lateral span of the industrial camera's field of view covers the maximum lateral travel of the laser rust removal operation area carried by the robot, and the vertical distance between the optical center of the industrial camera and the plane of the ship section is fixed.
[0013] Furthermore, in step S3, if the closure seam corrosion condition is identified, the continuous + pulse composite laser cleaning mode of the first power level is selected, and the laser head is controlled to perform three reciprocating scans and cleaning in the closure seam area. If the corrosion is identified as thick, select the continuous + pulsed composite laser cleaning mode of the second power level, and control the laser head to perform no more than two scanning cleanings. The second power level is higher than the first power level. If the condition is identified as light surface rust, select the continuous + pulsed composite laser cleaning mode of the third power level, and control the laser head to perform no more than two scanning cleanings. The third power level is lower than the first power level. If the condition is identified as not requiring rust removal, a lane-changing command is issued to the wall-climbing robot, controlling the robot to move to the next work point.
[0014] Furthermore, in step S5, the electric pan-tilt head is a two-degree-of-freedom electric pan-tilt head, and the industrial camera is fixed to the actuator of the electric pan-tilt head. The specific protective action of the industrial camera is as follows: Before the laser beam is emitted, the industrial camera is rotated and retracted from its fixed shooting position by an electric pan-tilt head, so that the camera lens and photosensitive element are completely facing away from the laser working area and the direction of laser reflection. After the laser cleaning operation stops, the industrial camera is driven by an electric pan-tilt head to rotate in the opposite direction and reset to a fixed shooting position.
[0015] Further, in step S6, the ROI area for the rust removal operation is determined. Specifically, determining the ROI area for the rust removal operation includes: Step S61) Perform morphological post-processing on the pixel-level segmentation mask of the rusted area to remove isolated noise points with an area smaller than a preset threshold, fill the holes inside the mask, and obtain a binarized operation mask. Step S62) Extract the boundary pixels of the binarized mask and determine the leftmost and rightmost pixels in the horizontal direction of the rust area; Step S63) Based on the pre-calibrated camera intrinsic parameters and the fixed distance between the camera and the segmented plane, the pixel boundary is converted into the actual physical size. Combined with the relative installation position of the industrial camera and the laser head, the horizontal working stroke of the laser head is calculated, and a laser scanning trajectory covering only the rusted area is generated.
[0016] Furthermore, in step S7, the rust removal effect is inspected online, specifically including: Step S71) After performing grayscale conversion, median filtering, Gaussian blurring, and noise reduction preprocessing on the ROI region of the cleaned work area image, perform two-dimensional discrete Fourier transform to calculate the average frequency band energy of the mid-low frequency band within the ROI region. Step S72) Compare the calculated average energy of the frequency band with the preset compliance threshold, which is calibrated based on the image of the same steel sample that meets the ship rust removal standard in the laboratory. Step S73) If the average energy of the frequency band is lower than the threshold, the rust removal is deemed to be up to standard, and a lane-changing command is sent to the wall-climbing robot; If the average energy of the frequency band is higher than the threshold, the rust removal is deemed substandard, triggering a re-cleaning operation, and steps S5-S7 are repeated.
[0017] Furthermore, the threshold for achieving the target is set to 0.1; For cleaning of non-seamed areas, the cleaning should not be repeated more than twice. For the closure seam area, a preset three-round reciprocating scan cleaning is performed, and no online inspection step is required after the cleaning is completed.
[0018] A visual inspection device for a large-area laser rust removal and wall-climbing robot for ship sections is provided to implement the aforementioned visual inspection method for the large-area laser rust removal and wall-climbing robot for ship sections. The device includes an image acquisition module, an image processing and working condition recognition module, a process parameter matching module, a lens protection module, a work planning module, a rust removal effect inspection module, a main control module, and a communication module. The image acquisition module includes an industrial camera and a two-degree-of-freedom electric gimbal. The industrial camera is mounted on the front end of the laser rust removal wall-climbing robot via the two-degree-of-freedom electric gimbal to acquire original images and real-time operation images of the ship section plane. The image processing and working condition recognition module is deployed on the host computer and has a built-in pre-trained and fine-tuned image segmentation model for pixel-level segmentation of rust areas and classification of working conditions in the acquired images. The process parameter matching module has a built-in preset laser cleaning process parameter library, which is used to match the corresponding laser cleaning mode and process parameters according to the identified working condition category. The lens protection module is connected to the two-degree-of-freedom electric gimbal drive and is used to control the pose switching of the industrial camera before and after laser cleaning operations to achieve protection of the lens and photosensitive element. The job planning module is used to calculate the working range of the laser head and generate the laser scanning trajectory based on the segmentation results of the rusted area; The rust removal effect inspection module is used to process and analyze the images after cleaning, complete the online inspection of the rust removal effect, and output the inspection results; The main control module is electrically connected to each of the above modules to coordinate the working timing and data transmission of each module; The communication module is used to realize communication connections between industrial cameras and host computers, host computers and wall-climbing robot control systems, and laser cleaning equipment.
[0019] Compared with the prior art, the present invention has the following advantages: By finely dividing the ship section planar operation objects into four types of working conditions, and combining the finely tuned SAM model, pixel-level accurate segmentation of rusted areas and adaptive identification of working conditions are achieved. Different laser cleaning process parameters are matched for different working conditions, which is different from the traditional fixed parameter full-coverage cleaning mode. This ensures rust removal quality, reduces energy consumption per unit area, avoids under-rust removal and over-rust removal, and protects the steel plate substrate from damage. Based on the visual segmentation results, a laser scanning trajectory is generated that only covers the rusted area, enabling selective rust removal operations, avoiding ineffective cleaning of non-rusted areas, and improving the working efficiency of the wall-climbing robot; A closed-loop negative feedback control system is constructed, which integrates rust detection, process matching, cleaning operations, camera protection, and real-time inspection. The system uses spectrum analysis to achieve online real-time inspection of rust removal effectiveness and automatically triggers re-cleaning operations for unqualified areas, reducing manual intervention and rework rates. Design a camera lens protection solution by adjusting the camera position using a gimbal so that the image sensor faces away from the laser working area, thus avoiding damage to the image sensor caused by high-intensity laser reflection light; This invention optimizes the special corrosion conditions of the joints between ship sections and matches a composite laser cleaning mode, solving the problem that the mixture of paint and rust in the joints is difficult to clean in one go, reducing dust pollution and improving the working environment. Attached Figure Description
[0020] Figure 1 This is a schematic diagram of the camera taking pictures according to the present invention; Figure 2 This is a schematic diagram of the camera of the present invention when it is retracted; Figure 3 This is a schematic diagram of the laser cleaning segmented planar visual inspection process of the present invention; Figure Labels 1. Wall-climbing robot; 2. Two-DOF motorized gimbal; 3. Industrial camera. 4 lenses, 5 ship section plane, 6 laser head, 7 sliding table. Detailed Implementation
[0021] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] This embodiment proposes a visual inspection method for a wall-climbing robot used for laser rust removal on large flat surfaces in ship sections. It is applied to secondary rust removal operations on the surface of the sections during ship section manufacturing, and is used in conjunction with a magnetically adsorbed wall-climbing robot 1 equipped with a composite laser cleaning device. Figure 3 As shown, the specific steps are as follows: Step S1) Collect nearly a thousand original images of typical shipyard sections with different lighting conditions and different degrees of corrosion. Manually label the areas of typical corrosion images, labeling positive samples and negative samples (including paint, stains, and manual chalk marks). Original images of the closure joint area of the ship section plane 5 were collected. The shape of the closure joint and the mixed paint and rust areas at the closure joint were individually labeled and studied to establish a raw database of typical rust images of the section.
[0023] Step S2) The wall-climbing robot 1 is equipped with a Yabo Intelligent K230 vision industrial camera 3. The industrial camera 3 is installed at the front end of the wall-climbing robot 1 via a two-degree-of-freedom electric gimbal 2. The field of view of the industrial camera 3 exceeds the longest horizontal stroke of the laser rust removal operation area carried by the robot. After the wall-climbing robot 1 travels along the preset path to the starting point of the work path, the industrial camera 3 is adjusted to a fixed shooting posture by the two-degree-of-freedom electric gimbal 2 to collect real-time images of the current ship section plane 5. The collected real-time section surface image data is transmitted to the host computer.
[0024] Step S3) The host computer runs a vision processing environment based on Python, which is pre-loaded with a SAM model fine-tuned by transfer learning. This model is optimized using the ship corrosion image dataset pre-annotated in step S1 to adapt it to the corrosion recognition and segmentation task of ship segment plane 5. After the real-time images are acquired and transmitted to the host computer, the rusted parts in the images are segmented at the pixel level using the SAM model. The segmentation mask of the rusted area is output. Based on the statistical characteristics of the mask area, four working conditions are automatically determined: "light surface rust", "thick rust", "seam rust" and "no rust removal required". A confidence score is also attached. Areas with a confidence score of more than 70% are determined to be features that require rust removal.
[0025] Step S4) Based on the working condition classification results output in step S3, the host computer queries the preset laser cleaning process parameter library and matches the corresponding laser cleaning mode and process parameters. The specific matching rules are as follows: If the segmented closure seam corrosion condition is identified, a continuous plus pulsed composite laser cleaning operation mode is selected. The matching parameters are: continuous laser power 4600W, scanning rate 18000mm / s; pulsed laser power 900W, frequency 1250KHz, pulse width 120ns, scanning rate 40000mm / s, maximum laser width 260mm. The laser head 6 is controlled to perform three reciprocating scans and cleaning in the closure seam corrosion area to ensure thorough removal of the paint and rust mixture at the closure seam location. If a thick rust condition is identified, a continuous + pulsed composite laser cleaning mode is matched with the following parameters: continuous laser power 5000W, scanning rate 12000mm / s; pulsed laser power 950W, frequency 700KHz, pulse width 350ns, scanning rate 35000mm / s, maximum laser width 215mm, and no more than two scans. If light surface rust is detected, a continuous + pulsed composite laser cleaning mode is selected, with the following parameters: continuous laser power 3500W, scanning rate 18000mm / s; pulsed laser power 1000W, frequency 700KHz, pulse width 350ns, scanning rate 35000mm / s, maximum laser width 215mm, and no more than two scans.
[0026] If no rust is detected on the current ship section plane 5, it is determined that no rust removal is required. The host computer sends an instruction to the wall-climbing robot 1, which can directly change course to the next location to prepare for taking a new photo without laser cleaning.
[0027] Step S5) Before starting laser cleaning, the host computer issues a command to drive the industrial camera 3 from the fixed shooting position to rotate and retract via the two-degree-of-freedom electric gimbal 2, so that the camera lens 4 and the photosensitive element are facing away from the laser working area and the direction of laser reflection, so as to avoid damage to the photosensitive element by the laser reflection light.
[0028] Step S6) Based on the pixel-level segmentation mask of the rust area output in step S3, the host computer first performs morphological post-processing to remove isolated noise points with an area smaller than a preset threshold, fill the holes inside the segmentation mask, and obtain a binarized operation mask. For the segmented rusted area, select the pixel boundary points of the mask, determine the specific ROI area that needs to be rusted, and extract the leftmost and rightmost pixels of the segmented rusted area. Based on the pre-calibrated camera intrinsic parameters, and combined with the fixed distance between the industrial camera 3 and the ship section plane 5, the actual physical dimensions of the rust boundary are obtained by proportionally converting the pixel value range. Based on the installation position of the industrial camera 3 fixed at the central axis of the wall-climbing robot 1, the horizontal cleaning stroke distance of the laser head 6 moving on the slide table 7 is calculated. After the laser head 6 reaches the leftmost end of its travel under the drive of the motor, it starts the rust removal operation by line scanning according to the process parameters matched in step S4, and performs scanning and cleaning on the rusted area to ensure stable working efficiency.
[0029] Step S7) After the laser head 6 completes a cleaning operation, it turns off the light output. The industrial camera 3 is reset to a fixed shooting position by the two-degree-of-freedom electric pan-tilt head 2, and the segmented surface images after laser rust removal are collected and sent to the host computer.
[0030] The host computer extracts the ROI region of the segmented surface image after laser cleaning, and performs grayscale conversion, median filtering, Gaussian blurring, and filtering noise reduction preprocessing in sequence. Then, it performs two-dimensional discrete Fourier transform to calculate the average energy of the low-frequency band in the ROI region. The low-frequency region corresponds to the overall brightness and gentle changes of the image, while the mid-frequency region corresponds to the texture details and medium-scale features in the image. The calculated average energy of the frequency band is compared with the average energy of the mid-low frequency band of the surface image of the same source section sample that has been pre-calibrated in the laboratory and meets the Sa2.5 grade standard of the classification society for steel plate pretreatment. In this embodiment, the threshold for human experience is set to 0.1, and the judgment and feedback rules are as follows: If the average energy of the ROI area after cleaning is less than 0.1, the rust removal is deemed to have met the standard. The host computer sends a lane change command to the wall-climbing robot 1, controlling the wall-climbing robot 1 to move to the next segment position for re-photographing. If the average energy of the frequency band is higher than 0.1, it is judged as substandard. The previous process parameters are used to perform a second cleaning operation. Then, the wall-climbing robot 1 changes course and goes to the next segment position. After the cleaning is completed, the effect is inspected again. In order to prevent laser damage to the steel plate substrate, the number of cleaning cycles for the non-joining seam area of the segment plane shall not exceed two.
[0031] For the segmented closure seam location, three repeated scanning operations are performed directly, without the need for photo inspection steps, and the wall-climbing robot 1 is directly controlled to change course to the next work point.
[0032] This embodiment provides a visual inspection device for implementing the above method, such as... Figure 1 and Figure 2 As shown, it includes: The image acquisition module includes a Yabo Intelligent K230 industrial camera 3 and a two-degree-of-freedom electric gimbal 2. The industrial camera 3 is mounted on the front end of the laser rust removal wall-climbing robot 1 via the two-degree-of-freedom electric gimbal 2, and is used to acquire original images and real-time operation images of the ship section plane 5. The calibration module is used to complete the calibration of the intrinsic and extrinsic parameters of the industrial camera 3, as well as the conversion calibration between pixel size and actual physical size; The image processing and working condition recognition module is deployed on the host computer. It has a built-in SAM model finely tuned with the ship corrosion dataset. It is used to perform pixel-level segmentation of the corrosion area and classification of working conditions in the acquired images, and outputs segmentation masks and working condition labels. The process parameter matching module has a built-in preset laser cleaning process parameter library, which stores parameters such as laser power, scanning speed, pulse frequency, pulse width, and number of scans corresponding to four types of working conditions. It matches the corresponding cleaning process based on the working condition identification results. The lens protection module is electrically connected to the drive motor of the two-degree-of-freedom electric gimbal 2. It is used to control the two-degree-of-freedom electric gimbal 2 to drive the industrial camera 3 to complete the pose switching according to the laser operation sequence, so as to protect the lens 4. The job planning module is used to perform pixel-to-physical size conversion based on the corrosion segmentation results, calculate the working stroke of the laser head 6, and generate a laser scanning trajectory that only covers the corrosion area. The rust removal effect inspection module has a built-in spectrum analysis unit, which is used to preprocess and Fourier transform the images after cleaning, calculate the average energy of the frequency band, and conduct online inspection and standard determination of the rust removal effect. The main control module is electrically connected to each of the above modules and is used to coordinate the working timing and data transmission of each module. The communication module is used to realize the communication connection between the industrial camera 3 and the host computer, the host computer and the motion control system of the wall-climbing robot 1, and the laser cleaning equipment; Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A visual inspection method for a wall-climbing robot used in large-area laser cleaning and rust removal of ship sections, characterized in that, Includes the following steps, Step S1) Establish a database of typical planar corrosion images of ship sections; Original images of planar corrosion of ship sections were collected, and the corrosion areas and joint areas in the images were manually labeled to construct a labeled dataset containing different corrosion types and different lighting conditions. Step S2) After the wall-climbing robot reaches the work site, it uses the onboard industrial camera to collect real-time planar images of the ship sections in the current work area; Step S3) The industrial camera performs rust segmentation and working condition identification on the real-time acquired images. Through the pre-trained and fine-tuned image segmentation model SAM model, the rust area in the image is segmented at the pixel level, and the working condition identification of the working area is divided into four categories: closure seam rust, thick rust, light surface rust, and no rust removal required. Step S4) Based on the working condition category identified by the industrial camera, match the preset laser cleaning process parameter library to match the corresponding laser cleaning mode and process parameters for different working conditions; Step S5) Before starting the laser cleaning operation, perform the industrial camera protection action by adjusting the position of the industrial camera with the electric pan-tilt head so that the photosensitive element of the industrial camera faces away from the laser operation area. Step S6) Based on the pixel-level segmentation results of the rusted area, determine the ROI area for the rust removal operation, calculate the actual working stroke range of the laser head, plan the laser scanning trajectory, and execute the laser cleaning operation. Step S7) After the laser cleaning operation is completed, the industrial camera is reset to the shooting position by controlling the electric pan-tilt head, and the image of the cleaned work area is collected to inspect the rust removal effect online. The inspection results are fed back to the wall-climbing robot control system to form a closed-loop control.
2. The visual inspection method according to claim 1, characterized in that, In step S1, the labeled dataset includes positive rust samples and negative samples of paint / stains / artificial chalk markings.
3. The visual inspection method according to claim 1, characterized in that, In step S2, an industrial camera is installed at the front end of the wall-climbing robot. The lateral span of the industrial camera's field of view covers the maximum lateral travel of the laser rust removal operation area carried by the robot. The vertical distance between the optical center of the industrial camera and the plane of the ship section is fixed each time a picture is taken.
4. The visual inspection method according to claim 1, characterized in that, In step S3, if the closure seam corrosion condition is identified, the first power level continuous + pulse composite laser cleaning mode is selected, and the laser head is controlled to perform three reciprocating scans and cleaning in the closure seam area. If the corrosion is identified as thick, select the continuous + pulsed composite laser cleaning mode of the second power level, and control the laser head to perform no more than two scanning cleanings. The second power level is higher than the first power level. If the condition is identified as light surface rust, select the continuous + pulsed composite laser cleaning mode of the third power level, and control the laser head to perform no more than two scanning cleanings. The third power level is lower than the first power level. If the condition is identified as not requiring rust removal, a lane-changing command is issued to the wall-climbing robot, controlling the robot to move to the next work point.
5. The visual inspection method according to claim 1, characterized in that, In step S5, the electric pan-tilt head is a two-degree-of-freedom electric pan-tilt head, and the industrial camera is fixed to the actuator of the electric pan-tilt head. The specific protective action of the industrial camera is as follows: Before the laser beam is emitted, the industrial camera is rotated and retracted from its fixed shooting position by an electric pan-tilt head, so that the camera lens and photosensitive element are completely facing away from the laser working area and the direction of laser reflection. After the laser cleaning operation stops, the industrial camera is driven by an electric pan-tilt head to rotate in the opposite direction and reset to a fixed shooting position.
6. The visual inspection method according to claim 1, characterized in that, In step S6, the ROI area for the rust removal operation is determined. Specifically, determining the ROI area for the rust removal operation includes: Step S61) Perform morphological post-processing on the pixel-level segmentation mask of the rusted area to remove isolated noise points with an area smaller than a preset threshold, fill the holes inside the mask, and obtain a binarized operation mask. Step S62) Extract the boundary pixels of the binarized mask and determine the leftmost and rightmost pixels in the horizontal direction of the rust area; Step S63) Based on the pre-calibrated camera intrinsic parameters and the fixed distance between the camera and the segmented plane, the pixel boundary is converted into the actual physical size. Combined with the relative installation position of the industrial camera and the laser head, the horizontal working stroke of the laser head is calculated, and a laser scanning trajectory covering only the rusted area is generated.
7. The visual inspection method according to claim 6, characterized in that, In step S7, the rust removal effect is inspected online, specifically including: Step S71) After performing grayscale conversion, median filtering, Gaussian blurring, and noise reduction preprocessing on the ROI region of the cleaned work area image, perform two-dimensional discrete Fourier transform to calculate the average frequency band energy of the mid-low frequency band within the ROI region. Step S72) Compare the calculated average energy of the frequency band with the preset compliance threshold, which is based on the calibration of the same source steel sample image in the laboratory that meets the ship rust removal Sa2.5 standard; Step S73) If the average energy of the frequency band is lower than the threshold, the rust removal is deemed to be up to standard, and a lane-changing command is sent to the wall-climbing robot; If the average energy of the frequency band is higher than the threshold, the rust removal is deemed substandard, triggering a re-cleaning operation, and steps S5-S7 are repeated.
8. The visual inspection method according to claim 7, characterized in that, The threshold for achieving the target is set to 0.1; For cleaning operations in non-sealing areas, in order to protect the steel plate substrate, the same area should not be cleaned more than twice. For the closure seam area, a preset three-round reciprocating scan cleaning is performed, and no online inspection step is required after the cleaning is completed.
9. A visual inspection device for a large-area laser rust removal climbing robot for ship sections, characterized in that, The visual inspection method for a large-area laser rust removal and wall-climbing robot for ship sections, as described in any one of claims 1-8, is used to implement the method described in claims 1-8. It includes an image acquisition module, an image processing and working condition recognition module, a process parameter matching module, a lens protection module, a job planning module, a rust removal effect inspection module, a main control module, and a communication module; The image acquisition module includes an industrial camera and a two-degree-of-freedom electric gimbal. The industrial camera is mounted on the front end of the laser rust removal wall-climbing robot via the two-degree-of-freedom electric gimbal to acquire original images and real-time operation images of the ship section plane. The image processing and working condition recognition module is deployed on the host computer and has a built-in pre-trained and fine-tuned image segmentation model for pixel-level segmentation of rust areas and classification of working conditions in the acquired images. The process parameter matching module has a built-in preset laser cleaning process parameter library, which is used to match the corresponding laser cleaning mode and process parameters according to the identified working condition category. The lens protection module is connected to the two-degree-of-freedom electric gimbal drive and is used to control the pose switching of the industrial camera before and after laser cleaning operations to achieve protection of the lens and photosensitive element. The job planning module is used to calculate the working range of the laser head and generate the laser scanning trajectory based on the segmentation results of the rusted area; The rust removal effect inspection module is used to process and analyze the images after cleaning, complete the online inspection of the rust removal effect, and output the inspection results; The main control module is electrically connected to each of the above modules to coordinate the working timing and data transmission of each module; The communication module is used to realize communication connections between industrial cameras and host computers, host computers and wall-climbing robot control systems, and laser cleaning equipment.