A square iron multi-face laser derusting system and method based on double-robot cooperation

Abstract

The application relates to the technical field of automation control, and discloses a square iron multi-surface laser rust removal system and method based on double-robot cooperation. The method places the square iron at a cleaning station through a carrying robot, a laser cleaning robot acquires the square iron pose by using a first vision module and dynamically plans a cleaning path, and laser cleaning is carried out; in the cleaning process, a second vision module monitors the cleaning degree and feeds back to a central control system, and the cleaning parameters are dynamically adjusted. After one surface is cleaned, the carrying robot overturns the square iron, the laser cleaning robot continues to clean other surfaces, and the cleaning of all surfaces is completed. The application realizes the automatic continuous cleaning of the square iron multi-surface, and improves the efficiency and cleaning quality consistency.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, specifically to a multi-faceted laser rust removal system and method for square iron based on dual-robot collaboration. Background Technology

[0002] Traditional rust removal of square iron relies heavily on manual operation or a single robot system, which suffers from problems such as low efficiency, poor precision, and high labor intensity.

[0003] Manual operation makes it difficult to guarantee consistent cleaning quality and requires highly skilled operators. Single-robot systems are limited by the range of motion of the robotic arm and the accessibility of the cleaning head, making it difficult to achieve continuous multi-faceted cleaning. Frequent manual intervention is required to adjust the workpiece position, resulting in low production efficiency. In addition, traditional methods lack a real-time feedback mechanism and cannot dynamically adjust cleaning parameters according to the degree of corrosion, which can easily lead to energy waste or incomplete cleaning.

[0004] With the increasing demand for industrial automation, existing technologies can no longer meet the requirements of efficient, precise, and intelligent production, and there is an urgent need for a new type of multi-faceted laser rust removal system. Summary of the Invention

[0005] To address the aforementioned technical shortcomings, the purpose of this invention is to provide a multi-faceted laser rust removal system and method for square iron based on dual-robot collaboration, thereby solving the problems of low efficiency, poor precision, and insufficient automation in the existing multi-faceted cleaning of square iron.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for multi-faceted laser rust removal of square iron based on dual-robot collaboration, the method comprising:

[0008] Step S1: The handling robot places the square iron on the positioning platform of the cleaning station;

[0009] Step S2: The laser cleaning robot acquires the positional data of the square iron through the first vision module installed at the end of its robotic arm, and dynamically plans the cleaning path based on the positional data to perform laser cleaning on the first surface of the square iron to be cleaned; during the cleaning process, the robot acquires the image of the first surface to be cleaned through the second vision module integrated next to the laser cleaning head to judge the degree of cleaning, and sends the feedback information to the central control system, which dynamically adjusts the laser cleaning parameters.

[0010] Step S3: After the first surface to be cleaned is cleaned, the laser cleaning robot moves to a safe position; after the central control system confirms that the cleaning station is safe, it instructs the transport robot to enter the station, grab and flip the square iron so that the second surface to be cleaned faces upward and put it back on the positioning table.

[0011] Step S4: After the central control system confirms the safety of the workstation again, it instructs the laser cleaning robot to enter the workstation; the first vision module performs pose recognition on the returned square iron, corrects the cleaning path, and then repeats the cleaning process in step S2 to clean the second surface to be cleaned.

[0012] Step S5: For the remaining surfaces of the square iron to be cleaned, repeat steps S3 and S4 until all predetermined surfaces are cleaned, and then the handling robot will transfer the square iron to the unloading area.

[0013] Preferably, in one possible implementation of the first aspect, the process of acquiring the square iron pose data through the first vision module includes:

[0014] The camera captures multi-angle image sequences of the iron on the positioning platform, and the image processing unit performs multi-scale feature extraction to obtain a set of geometric feature points including corner points and edge contours;

[0015] The geometric feature point set is registered with the pre-stored 3D model of the square iron. A pose estimation algorithm based on deep learning is used to calculate the six-degree-of-freedom pose of the square iron relative to the positioning stage. This pose includes position coordinates and rotation angle.

[0016] The six-degree-of-freedom pose data is transmitted to the central control system via a real-time communication interface.

[0017] Preferably, in one possible implementation of the first aspect, the dynamic programming cleaning path includes:

[0018] The central control system uses a six-degree-of-freedom pose to call a pre-planned baseline cleaning path;

[0019] Based on the deviation between the six-degree-of-freedom pose and the preset standard pose, a trajectory interpolation algorithm is used to adjust the coordinates and attitude of the trajectory points of the reference cleaning path.

[0020] Preferably, in one possible implementation of the first aspect, the process of determining the degree of cleaning through the second vision module includes:

[0021] The second vision module periodically acquires images of the square iron surface during the laser cleaning process;

[0022] The image analysis algorithm converts the image from the RGB color space to the HSV color space and extracts the texture features of the surface. Based on the saturation component values ​​and texture features, the threshold segmentation method is used to distinguish between rusted areas and clean areas.

[0023] Calculate the proportion of pixels in the rusted area to the total number of pixels on the surface to be cleaned. When this proportion is lower than the preset cleanliness threshold, the current area is considered to be cleaned.

[0024] The cleaning feedback information, including a rust distribution map and cleaning progress percentage, is sent to the central control system.

[0025] Preferably, in one possible implementation of the first aspect, the image analysis algorithm employs a convolutional neural network model to achieve pixel-level segmentation of the rusted area;

[0026] A convolutional neural network model consists of an encoder and a decoder structure. The encoder is used to extract image features, and the decoder is used to output a segmentation map of the rusted region with the same resolution as the input image.

[0027] Preferably, in one possible implementation of the first aspect, the process of dynamically adjusting the laser cleaning parameters includes:

[0028] The central control system receives cleaning feedback information and analyzes the deviation between the current rust coverage and the target value.

[0029] The deviation is input to the fuzzy logic controller, which outputs the adjustment amount of laser power density, scanning speed and pulse frequency according to the preset rule base;

[0030] Parameter adjustments are performed based on the cleaning parameter mapping table stored in the central control system, forming an adaptive closed-loop control.

[0031] Preferably, in one possible implementation of the first aspect, the confirmation of the cleaning station safety includes a space safety interlock and a laser safety interlock;

[0032] The space safety interlock is to confirm that the robotic arm of the laser cleaning robot has fully moved to the preset safe waiting position and that its motion mechanism is in a locked state.

[0033] The laser safety interlock is established to confirm that the laser's light output enable signal has been disconnected, and at the same time, the physical protective door of the cleaning station closes, and the sensor feeds back a valid signal.

[0034] Preferably, in one possible implementation of the first aspect, the pose recognition and correction of the cleaning path includes:

[0035] The first vision module re-acquires images and calculates new six-degree-of-freedom poses after the square iron is placed back on the positioning stage.

[0036] The central control system compares the new pose with the historical pose before the flip and calculates the pose transformation matrix;

[0037] The transformation matrix is ​​used to perform coordinate transformation on the cleaning path of the previous surface to be cleaned, generating a corrected cleaning path suitable for the current surface to be cleaned.

[0038] Preferably, in one possible implementation of the first aspect, the method further includes using a sequence controller to control the cleaning sequence;

[0039] The sequence controller records the total number of square iron surfaces to be cleaned and the number of surfaces that have been cleaned.

[0040] After each surface is cleaned, the sequence controller automatically triggers the transport robot to perform a flipping operation and instructs the laser cleaning robot to prepare to clean the next surface, until all recorded surfaces have been cleaned.

[0041] Secondly, the present invention provides a multi-faceted laser rust removal system for square iron based on dual-robot collaboration. The system is used to implement the multi-faceted laser rust removal method for square iron based on dual-robot collaboration as described in the first aspect, comprising:

[0042] The central control system is used to control the rust removal process;

[0043] The cleaning station is equipped with a positioning platform for holding square iron pieces.

[0044] The handling robot communicates with the central control system and is used to transfer square irons between pallets, cleaning stations and material boxes, and to perform handling and flipping operations on the square irons.

[0045] The laser cleaning robot is connected to the central control system, and a laser cleaning head is installed at the end of its robotic arm.

[0046] The first vision module is fixedly installed at the end of the robotic arm of the laser cleaning robot and connected to the central control system to identify the position and posture of the square iron.

[0047] The second vision module, integrated next to the laser cleaning head and connected to the central control system, is used to monitor the cleaning process.

[0048] The beneficial effects of this invention are as follows: This invention achieves automated continuous cleaning of multiple surfaces of square iron through the collaborative operation of two robots, thereby improving production efficiency.

[0049] The integrated application of the first vision module and the second vision module, combined with deep learning algorithms and fuzzy control technology, enables accurate pose recognition, dynamic planning of cleaning paths, and adaptive adjustment of parameters, ensuring consistent cleaning quality.

[0050] Space safety interlock and laser safety interlock mechanisms ensure the safety of equipment operation.

[0051] Sequence controllers simplify the operation process and reduce the need for manual intervention.

[0052] The system optimizes energy utilization and reduces cleaning agent consumption through closed-loop control, which aligns with the concept of green manufacturing and has significant economic and social benefits. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 This application provides a flowchart of a method for multi-faceted laser rust removal of square iron based on dual-robot collaboration.

[0055] Figure 2 This application provides a detailed step diagram of a method for multi-faceted laser rust removal of square iron based on dual-robot collaboration. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0057] Example 1: As Figure 1 and Figure 2 As shown, this invention provides a method for multi-faceted laser rust removal of square iron based on dual-robot collaboration, comprising:

[0058] Step S1: The transport robot places the square iron on the positioning platform of the cleaning station.

[0059] In this embodiment, the handling robot is a six-axis industrial robot equipped with a torque-controllable gripper for reliably grasping square iron. After the handling robot picks up the rusty square iron from the pallet, the central control system instructs it to move along a preset trajectory to the cleaning station.

[0060] The cleaning station is located at the intersection of the working areas of the transport robot and the laser cleaning robot, and is equipped with a square iron positioning platform for placing square iron pieces. The central control system, with a programmable logic controller (PLC) at its core, communicates with the transport robot controller to control the transport robot to perform the placement actions.

[0061] During movement, the transport robot relies on its positioning to ensure that the square iron is placed stably in the center area of ​​the positioning platform with the first surface to be cleaned facing upwards. Before the placement action is executed, the central control system confirms through a spatial safety interlock mechanism that the laser cleaning robot has completely exited the core area of ​​the cleaning station and is in a safe waiting position, with its motion mechanism locked to prevent spatial interference between the two robots.

[0062] Step S2: The laser cleaning robot acquires the positional data of the square iron through the first vision module installed at the end of its robotic arm, and dynamically plans the cleaning path based on the positional data to perform laser cleaning on the first surface of the square iron to be cleaned. During the cleaning process, the robot acquires the image of the first surface to be cleaned through the second vision module integrated next to the laser cleaning head to judge the degree of cleaning, and sends the feedback information to the central control system, which dynamically adjusts the laser cleaning parameters.

[0063] In this embodiment, after the laser cleaning robot is started, it acquires the pose of the square iron through a first vision module installed at the end of its robotic arm. The first vision module includes a high-resolution industrial camera, which acquires multi-angle image sequences of the square iron on the positioning platform. The image processing unit performs multi-scale feature extraction on the acquired images and uses the Gaussian pyramid algorithm to generate images at different scales. The Gaussian pyramid algorithm starts from the original image and constructs a series of image layers with decreasing resolution through progressive downsampling to capture feature details at different scales.

[0064] Then, the Harris corner detector is used to extract corner features. The Harris corner detector locates corner positions by analyzing gradient changes in local regions of the image. Corners are characterized by significant gradient changes in multiple directions. Simultaneously, the Canny edge detector is used to extract edge contours. The Canny edge detector includes gradient calculation, non-maximum suppression, and double thresholding steps to generate coherent and refined edge lines. These feature point sets together constitute the geometric feature descriptor.

[0065] These geometric feature point sets are registered with a pre-stored 3D model of a square iron frame, which is a mesh model generated from CAD data. The registration process uses an iterative nearest-point algorithm for initial alignment. The iterative nearest-point algorithm optimizes the transformation matrix iteratively. In each iteration, the distance between the feature point and the nearest point on the model surface is calculated, and the overall distance error is minimized. The rotation and translation transformations are gradually adjusted until convergence to the optimal registration state.

[0066] Then, a deep learning-based pose estimation algorithm is used. This algorithm is based on a convolutional neural network structure, with the registered feature data as the network input. High-level features are extracted progressively through multiple convolutional and pooling layers, ultimately outputting the six-DOF pose of the square iron relative to the positioning stage, including position coordinates and rotation angles. The six-DOF pose data is transmitted to the central control system via a real-time communication interface.

[0067] After receiving the pose data, the central control system dynamically plans the cleaning path. The system calls upon a pre-planned baseline cleaning path, which is a laser scanning trajectory designed for a standard pose. Based on the deviation between the actual six-degree-of-freedom pose and the preset standard pose, a trajectory interpolation algorithm is used to adjust the coordinates and attitude of the trajectory points on the baseline cleaning path.

[0068] The trajectory interpolation algorithm uses linear interpolation to compensate for positional deviations, which smooths the path by uniformly inserting intermediate points between the start and end points. It also uses quaternion interpolation to smoothly adjust rotational deviations, leveraging the spherical linear interpolation properties of quaternions to ensure the smoothness and continuity of rotational changes. These interpolation operations generate an optimized clean path that adapts to the actual pose.

[0069] The laser cleaning robot, following an adjusted cleaning path, controls the laser cleaning head to perform laser cleaning on the first surface of the square iron. The laser cleaning head is connected to a laser generator via fiber optic cable, outputting continuous-wave laser light with adjustable power. During the cleaning process, a second vision module integrated next to the laser cleaning head periodically acquires images of the square iron surface. The second vision module includes a color CCD camera equipped with a near-infrared filter to enhance corrosion contrast.

[0070] The acquired image is processed using image analysis algorithms. First, the image is converted from the RGB color space to the HSV color space. The color space conversion algorithm enhances the contrast of the rusted area by calculating the hue, saturation, and brightness components. Then, the saturation component and texture features are extracted. The texture features are used to calculate the contrast and entropy values ​​using the gray-level co-occurrence matrix algorithm. The gray-level co-occurrence matrix algorithm statistically analyzes the spatial relationship of image pixel pairs to generate texture indices to quantify surface irregularities.

[0071] Based on saturation component values ​​and texture features, a threshold segmentation method is used to distinguish between rusted and clean areas. The threshold segmentation algorithm dynamically sets the segmentation threshold, which is adaptively adjusted based on historical image data. The proportion of pixels in the rusted area to the total pixels of the surface to be cleaned is calculated, and when this proportion is lower than a preset cleanliness threshold, the current area is considered to have been cleaned successfully.

[0072] In image analysis algorithms, threshold segmentation employs a convolutional neural network (CNN) model to achieve pixel-level segmentation of rusted regions. The CNN model comprises an encoder and a decoder. The encoder consists of multiple convolutional and pooling layers. The convolutional layers use a sliding window to extract local features, while the pooling layers reduce the feature map size and retain key information for feature extraction. The decoder consists of deconvolutional and upsampling layers. The deconvolutional layers progressively restore spatial details, and the upsampling layers enlarge the feature map to its original resolution, outputting a rusted region segmentation map with the same resolution as the input image. Training data includes labeled rusted images, and optimization is performed using a cross-entropy loss function. The training process adjusts network weights through backpropagation to minimize segmentation error.

[0073] The cleaning feedback information, including a rust distribution map and a cleaning progress percentage, is sent to the central control system via Ethernet. Upon receiving the feedback information, the central control system analyzes the deviation between the current rust coverage and the target value. This deviation is input to a fuzzy logic controller, which fuzzifies the precise deviation value into linguistic variables. In this embodiment, the linguistic variables include positive large, positive small, zero, negative small, and negative large. Then, a rule base based on expert experience is applied. The rule base uses if statements to define the input-output relationship, and finally, the output adjustment amount is defuzzified.

[0074] The fuzzy logic controller outputs adjustments to the laser power density, scanning speed, and pulse frequency based on a preset rule base. The rule base is designed based on expert experience. In this embodiment, the rule base includes: if the deviation is large positive, significantly increase the laser power density, significantly decrease the scanning speed, and significantly increase the pulse frequency; if the deviation is small positive, slightly increase the laser power density, slightly decrease the scanning speed, and slightly increase the pulse frequency; if the deviation is zero, the laser power density, scanning speed, and pulse frequency remain unchanged; if the deviation is small negative, slightly decrease the laser power density, slightly increase the scanning speed, and slightly decrease the pulse frequency; if the deviation is large negative, significantly decrease the laser power density, significantly increase the scanning speed, and significantly decrease the pulse frequency. These adjustments to the laser power density, scanning speed, and pulse frequency are based on preset parameters and do not exceed the maximum range of these parameters. Parameter adjustments are performed based on a cleaning parameter mapping table stored in the central control system. This mapping table associates the deviation range with the operation command, forming an adaptive closed-loop control to ensure consistent cleaning quality.

[0075] Step S3: After the first surface to be cleaned is cleaned, the laser cleaning robot moves to a safe position. After confirming that the cleaning station is safe, the central control system instructs the transport robot to enter the station, grab and flip the square iron so that the second surface to be cleaned faces upward and put it back on the positioning table.

[0076] In this embodiment, after the laser cleaning robot completes the cleaning of the first surface, its controller executes a pre-programmed retreat trajectory, driving the robotic arm to move along an optimized path to a preset safe waiting position. This safe waiting position is located outside the cleaning station, ensuring that the robotic arm is completely detached from the work area and avoiding spatial interference with the handling robot. Upon reaching the safe position, the laser cleaning robot's motion mechanism automatically enters a locked state, maintaining its position through the braking mechanism of the servo driver.

[0077] The central control system monitors the safety status of the cleaning station in real time. The safety confirmation process follows a dual composite interlocking mechanism, including spatial safety interlocking and laser safety interlocking. Spatial safety interlocking confirms that the laser cleaning robot's robotic arm has fully moved to the preset safe waiting position and that its motion mechanism is locked. Spatial safety interlocking first verifies the laser cleaning robot's status. The system reads real-time position feedback data from the laser cleaning robot controller via an Ethernet communication interface and compares the deviation between the actual coordinates of the robotic arm's end effector and the preset safe coordinates. Simultaneously, when the laser cleaning robot controller returns a ready signal, it determines that the motion mechanism is locked. The spatial safety interlocking condition is met when the deviation between the actual coordinates of the robotic arm's end effector and the preset safe coordinates is below a threshold and the laser cleaning robot controller returns a ready signal.

[0078] The laser safety interlock operates synchronously, confirming that the laser emission enable signal has been disconnected and that the physical safety door closure sensor at the cleaning station has provided a valid feedback signal. The central control system sends a query command to the laser generator to confirm that the laser emission enable signal is disconnected, and simultaneously detects the feedback signal from the physical safety door closure sensor at the cleaning station. The safety door sensor uses magnetic induction; it outputs a high-level valid signal when the door is fully closed. The laser safety interlock condition is met when both the laser emission enable signal and the physical safety door closure sensor at the cleaning station provide a valid feedback signal.

[0079] The central control system integrates the input signals of the space safety interlock and the laser safety interlock. The cleaning station is deemed to be in a safe state only when both the conditions of the space safety interlock and the laser safety interlock are met, that is, when the signals of the space safety interlock and the laser safety interlock input to the central control system are both valid.

[0080] After safety confirmation, the central control system sends a sequence of commands to the handling robot controller via the industrial Ethernet protocol. The handling robot controller parses the commands and drives the six-axis robotic arm along a pre-planned path into the cleaning station. The gripper at the end of the robotic arm features a torque-controllable design, with the gripping force adjusted via a pneumatic servo system. Once the gripper contacts the square block, a torque sensor provides real-time pressure data, and the controller dynamically adjusts the gripping force to a preset range for stable grasping.

[0081] After grasping, the handling robot performs a flipping motion. The flipping trajectory is pre-generated based on offline programming software, rotating around the long axis of the square iron by a specific angle. In this embodiment, the rotation angle is set to 90 degrees, ensuring that the adjacent surface to be cleaned faces upwards. The flipping path of the robotic arm in the air is calculated using an inverse kinematics algorithm to solve for the angular displacement of each joint, avoiding singularities and optimizing energy consumption. After flipping, the handling robot places the square iron back onto the positioning platform. Combining historical pose data from the vision system, it ensures that the second surface to be cleaned on the square iron faces upwards. After the placement action is completed, the gripper releases the square iron, and the handling robot exits the workstation to the waiting area.

[0082] Step S4: After the central control system reconfirms the safety of the workstation, it instructs the laser cleaning robot to enter the workstation; the first vision module performs pose recognition on the returned square iron, corrects the cleaning path, and then repeats the cleaning process in step S2 to clean the second surface to be cleaned.

[0083] In this embodiment, the central control system executes the workstation safety verification process again. This verification process is based on the interlocking mechanism in step S3, including dual verification of spatial safety interlock and laser safety interlock. After the safety verification is completed, the central control system sends a sequence of instructions to the laser cleaning robot controller via the industrial Ethernet protocol, driving the laser cleaning robot to enter the cleaning station along the pre-planned trajectory.

[0084] The first vision module initiates the pose recognition and correction procedure. First, it uses the pose acquisition procedure from step S2 to obtain pose data containing three-dimensional position coordinates and three rotation angles. Upon receiving the new six-degree-of-freedom pose, the central control system performs path correction.

[0085] The system compares the new pose with the historical pose stored before the flip and calculates the pose transformation matrix. The transformation matrix is ​​derived through homogeneous coordinate transformation and includes rotation matrices and translation vectors. Specifically, the Rodrigues formula is used to convert rotation angles into rotation matrices, which are then combined with translation vectors to construct a complete 4×4 transformation matrix. Using this transformation matrix, the reference cleaning path of the previous surface to be cleaned is transformed in coordinates; the reference cleaning path is a pre-planned laser scanning trajectory for a standard pose, stored as a series of trajectory point coordinates and orientations.

[0086] The coordinate transformation process applies the principle of rigid body transformation, multiplying the coordinates of each trajectory point by a transformation matrix to achieve rotation and translation mapping of the coordinate system, generating a corrected cleaning path suitable for the current surface to be cleaned.

[0087] After calibration, the laser cleaning robot performs the cleaning process described in step S2 based on the calibrated path to perform laser cleaning on the second surface to be cleaned.

[0088] Step S5: For the remaining surfaces of the square iron to be cleaned, repeat steps S3 and S4 until all predetermined surfaces are cleaned, and then the handling robot will transfer the square iron to the unloading area.

[0089] In this embodiment, the system uses a sequence controller to automate the sequential processing of multiple surfaces of the square iron to be cleaned. During system initialization, the sequence controller preloads the total number of surfaces to be cleaned on the square iron; in this embodiment, this number is set to four sides. During operation, a counter for surfaces that have been cleaned is maintained in real time.

[0090] After completing the laser cleaning operation on each surface to be cleaned, the laser cleaning robot sends a completion signal to the central control system. The central control system interprets the signal and instructs the sequence controller to increment the counter value. The controller then compares the number of completed surfaces with the total number of surfaces. If the values ​​are not equal, it automatically triggers the transport robot to perform a flipping operation and re-enters the process of steps S3 and S4; at the same time, it sends a preparation command to the laser cleaning robot, putting it into standby mode to wait for the cleaning task of the next surface.

[0091] The cycle continues until the sequence controller detects that the number of completed surfaces equals the total number of surfaces. At this point, the controller determines that all predetermined surfaces have been cleaned and sends a termination signal to the central control system. The central control system instructs the transport robot to pick up the cleaned square iron and move it to a dedicated material box in the unloading area. During the unloading process, the transport robot uses torque-controlled grippers to ensure the square iron is placed stably. After the square iron is neatly stacked, the system updates the production log and resets the status, preparing to process the next square iron, achieving fully automated continuous production.

[0092] Example 2: This invention provides a method for multi-faceted laser rust removal of square iron based on dual-robot collaboration. It further includes pre-identifying the square iron using a third vision module integrated above the pallet before the handling robot performs its material handling operation. This module uses a high-resolution camera to acquire images of the square iron surface, analyzes the rust distribution level and geometric dimensions using a convolutional neural network model, and transmits the identification data to the central control system in real time.

[0093] Based on the pre-identification results, the central control system dynamically pre-configures laser cleaning parameters, including initial power density and scanning speed, and optimizes the gripping trajectory and flipping sequence of the handling robot to improve overall collaborative efficiency. During subsequent cleaning, the pre-processed data is combined with real-time feedback from the second vision module to ensure that square iron pieces with different degrees of corrosion achieve consistent cleaning quality.

[0094] Example 3: This invention provides a multi-faceted laser rust removal system for square iron based on dual-robot collaboration, comprising:

[0095] The central control system is used to control the rust removal process;

[0096] The cleaning station is equipped with a positioning platform for holding square iron pieces.

[0097] The handling robot communicates with the central control system and is used to transfer square irons between pallets, cleaning stations and material boxes, and to perform handling and flipping operations on the square irons.

[0098] The laser cleaning robot is connected to the central control system, and a laser cleaning head is installed at the end of its robotic arm.

[0099] The first vision module is fixedly installed at the end of the robotic arm of the laser cleaning robot and connected to the central control system to identify the position and posture of the square iron.

[0100] The second vision module, integrated next to the laser cleaning head and connected to the central control system, is used to monitor the cleaning process.

[0101] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for multi-faceted laser rust removal of square iron based on dual-robot collaboration, characterized in that, The method includes: Step S1: The handling robot places the square iron on the positioning platform of the cleaning station; Step S2: The laser cleaning robot acquires the positional data of the square iron through the first vision module installed at the end of its robotic arm, and dynamically plans the cleaning path based on the positional data to perform laser cleaning on the first surface of the square iron to be cleaned; during the cleaning process, the robot acquires the image of the first surface to be cleaned through the second vision module integrated next to the laser cleaning head to judge the degree of cleaning, and sends the feedback information to the central control system, which dynamically adjusts the laser cleaning parameters. The process of determining the degree of cleaning using the second vision module includes: The second vision module periodically acquires images of the square iron surface during the laser cleaning process; The image analysis algorithm converts the image from the RGB color space to the HSV color space and extracts the texture features of the surface. Based on the saturation component values ​​and texture features, the threshold segmentation method is used to distinguish between rusted areas and clean areas. Calculate the proportion of pixels in the rusted area to the total number of pixels on the surface to be cleaned. When this proportion is lower than the preset cleanliness threshold, the current area is considered to be cleaned. The cleaning feedback information includes a rust distribution map and a percentage of cleaning progress, and is sent to the central control system. The image analysis algorithm uses a convolutional neural network model to achieve pixel-level segmentation of the rusted area; The convolutional neural network model includes an encoder and a decoder structure. The encoder is used to extract image features, and the decoder is used to output a segmentation map of the rusted region with the same resolution as the input image. Step S3: After the first surface to be cleaned is cleaned, the laser cleaning robot moves to a safe position; after the central control system confirms that the cleaning station is safe, it instructs the transport robot to enter the station, grab and flip the square iron so that the second surface to be cleaned faces upward and put it back on the positioning table. The confirmation of safety at the cleaning station includes spatial safety interlocking and laser safety interlocking; The space safety interlock is to confirm that the robotic arm of the laser cleaning robot has fully moved to the preset safe waiting position and that its motion mechanism is in a locked state. The laser safety interlock is to confirm that the laser output enable signal has been disconnected, and at the same time, the physical protective door of the cleaning station closes, and the sensor feeds back a valid signal. Step S4: After the central control system confirms the safety of the workstation again, it instructs the laser cleaning robot to enter the workstation; the first vision module performs pose recognition on the returned square iron, corrects the cleaning path, and then repeats the cleaning process in step S2 to clean the second surface to be cleaned. The pose recognition and correction cleaning path includes: The first vision module re-acquires images and calculates new six-degree-of-freedom poses after the square iron is placed back on the positioning stage. The central control system compares the new pose with the historical pose before the flip and calculates the pose transformation matrix; The transformation matrix is ​​used to perform coordinate transformation on the cleaning path of the previous surface to be cleaned, and a corrected cleaning path suitable for the current surface to be cleaned is generated. Step S5: For the remaining surfaces of the square iron to be cleaned, repeat steps S3 and S4 until all predetermined surfaces are cleaned, and then the handling robot will transfer the square iron to the unloading area.

2. The method for multi-faceted laser rust removal of square iron based on dual-robot collaboration as described in claim 1, characterized in that, The process of acquiring the pose data of the square iron through the first vision module includes: The camera captures multi-angle image sequences of the iron on the positioning platform, and the image processing unit performs multi-scale feature extraction to obtain a set of geometric feature points including corner points and edge contours; The geometric feature point set is registered with the pre-stored 3D model of the square iron. A pose estimation algorithm based on deep learning is used to calculate the six-degree-of-freedom pose of the square iron relative to the positioning stage. This pose includes position coordinates and rotation angle. The six-degree-of-freedom pose data is transmitted to the central control system via a real-time communication interface.

3. The method for multi-faceted laser rust removal of square iron based on dual-robot collaboration as described in claim 2, characterized in that, The dynamic programming cleaning path includes: The central control system uses a six-degree-of-freedom pose to call a pre-planned baseline cleaning path; Based on the deviation between the six-degree-of-freedom pose and the preset standard pose, a trajectory interpolation algorithm is used to adjust the coordinates and attitude of the trajectory points of the reference cleaning path.

4. The method for multi-faceted laser rust removal of square iron based on dual-robot collaboration as described in claim 1, characterized in that, The process of dynamically adjusting laser cleaning parameters includes: The central control system receives cleaning feedback information and analyzes the deviation between the current rust coverage and the target value. The deviation is input to the fuzzy logic controller, which outputs the adjustment amount of laser power density, scanning speed and pulse frequency according to the preset rule base; Parameter adjustments are performed based on the cleaning parameter mapping table stored in the central control system, forming an adaptive closed-loop control.

5. The method for multi-faceted laser rust removal of square iron based on dual-robot collaboration as described in claim 1, characterized in that, The method also includes using a sequence controller to control the cleaning sequence; The sequence controller records the total number of square iron surfaces to be cleaned and the number of surfaces that have been cleaned. After each surface is cleaned, the sequence controller automatically triggers the transport robot to perform a flipping operation and instructs the laser cleaning robot to prepare to clean the next surface, until all recorded surfaces have been cleaned.

6. A multi-faceted laser rust removal system for square iron based on dual-robot collaboration, characterized in that, The system is used to implement a method for multi-faceted laser rust removal of square iron based on dual-robot collaboration as described in any one of claims 1 to 5, comprising: The central control system is used to control the rust removal process; The cleaning station is equipped with a positioning platform for holding square iron pieces. The handling robot communicates with the central control system and is used to transfer square irons between pallets, cleaning stations and material boxes, and to perform handling and flipping operations on the square irons. The laser cleaning robot is connected to the central control system, and a laser cleaning head is installed at the end of its robotic arm. The first vision module is fixedly installed at the end of the robotic arm of the laser cleaning robot and connected to the central control system to identify the position and posture of the square iron. The second vision module, integrated next to the laser cleaning head and connected to the central control system, is used to monitor the cleaning process.

Images