A laser cutting machine adaptive sorting and discharging method and system

Abstract

This application provides an adaptive sorting and unloading method and system for laser cutting machines, belonging to the field of laser cutting vision sorting and inspection technology. The method includes: setting up a vision inspection device, performing vision calibration, and establishing a transformation relationship between the vision and robotic arm coordinate systems; analyzing the cutting program to determine the theoretical position and geometric features of the parts, optimizing the part gripping sequence using a genetic algorithm to complete coarse positioning; establishing a suction cup parameter library and matching the optimal suction cup combination based on the part's geometric features; generating motion commands based on the coarse positioning to control the robotic arm to move to the target position, acquiring the marker point image of the part, determining the search area and acquiring images, and performing image processing to accurately identify the part's position; calculating the precise coordinates of the part's gripping point in the robotic arm coordinate system based on the transformation relationship and the part's precise position, and controlling the robotic arm to execute unloading based on the part's theoretical position and geometric features. This application uses GCode analysis to optimize the path, performs precise vision positioning, and adaptively matches suction cups for unloading.

Description

Technical Field

[0001] This application belongs to the field of laser cutting vision sorting and inspection technology, specifically relating to an adaptive sorting and unloading method and system for laser cutting machines. Background Technology

[0002] Laser cutting technology is widely used in manufacturing due to its high precision and efficiency. However, after the laser cutting machine completes the cutting of metal sheets, how to efficiently and accurately sort and unload numerous parts of different shapes and sizes from the cutting table has become a key issue restricting the automation and intelligence level of the entire production process.

[0003] Traditional methods primarily rely on manual sorting and unloading, which has the following drawbacks: First, it is labor-intensive and inefficient, especially for cutting large batches of small parts, where prolonged manual operation easily leads to fatigue and errors. Second, manual operation is difficult to seamlessly integrate with subsequent processes, resulting in unstable production cycles. Although some automated sorting devices have emerged, they generally suffer from insufficient adaptability. For example, they struggle to handle parts of different sizes and contours, their sorting accuracy is limited by the repeatability of the robotic arm, failing to meet high-precision requirements, and they lack effective collision warning and protection mechanisms, easily damaging expensive suction cups or workpieces during the gripping process. In particular, existing solutions often fail to effectively utilize prior information from the cutting process, resulting in a lack of foresight and global optimization in the sorting process, and a low level of intelligence in the entire system.

[0004] Therefore, there is an urgent need for an intelligent sorting and unloading method that can adapt to different part characteristics, integrate visual information and cutting program data, and has safety protection capabilities, in order to solve the efficiency, accuracy and adaptability problems existing in the current technology. Summary of the Invention

[0005] In a first aspect, embodiments of this application provide an adaptive sorting and unloading method for a laser cutting machine, comprising the following steps: S1. Install a vision inspection device on the sorting equipment and perform vision calibration to establish the transformation relationship between the vision coordinate system and the robot arm base coordinate system; S2. Read and parse the GCode of the cutting program to obtain the theoretical position information and geometric features of each part, and use a genetic algorithm to optimize the gripping order of the parts and generate a coarse positioning coordinate sequence; S3. Establish a database containing parameters of each suction cup, and based on the geometric features of the parts, match and output the optimal suction cup combination required to grip each part from the database; S4. The host computer generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies the preset marker points on the part through the vision inspection device to obtain the image coordinates of the marker points; S5. Acquire images of the target area through a vision inspection device, perform image processing to accurately identify the position of the part, and simultaneously detect whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence signal of the part; S6. Based on the transformation relationship between the visual coordinate system and the robot arm base coordinate system and the precise coordinates of the parts, calculate the precise coordinates of the gripping points of each part in the robot arm base coordinate system; S7. Based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part, control the robotic arm to perform the unloading action.

[0006] Furthermore, the specific steps of step S1 are as follows: S11. Based on the gripping motion direction of the sorting equipment, a vision inspection device is installed on a suction cup bracket. The vision inspection device includes a camera, a ring-shaped infrared LED light source, and a diffuser plate. S12. Using the front crossbeam frame and the left crossbeam frame of the laser cutting machine table as a reference, adjust the installation height and shooting angle of the vision inspection device through the calibration tool so that the field of view of the vision inspection device covers the entire laser cutting machine table. S13. Adjust the illumination angle of the infrared light source so that the optical axis of the infrared light source deflects at a preset angle away from the baseline perpendicular to the laser cutting machine table, and the light is evenly diffused by the diffuser plate. S14. Obtain camera parameters and installation area range, calculate the theoretically recommended installation parameters required to cover the target monitoring area, compare the actual installation parameters with the theoretically recommended installation parameters, complete the installation compliance verification, and finally establish the transformation matrix between the visual coordinate system and the robotic arm base coordinate system.

[0007] Furthermore, the specific steps of step S2 are as follows: S21. Load the GCode file, open and read the file content using the ifstream class, parse the string stream into a JSON object using the nlohmann library, access the fields of the JSON object through key-value pairs, and extract the theoretical position information, geometric features, and gripping mode information of each part in the laser cutting machine coordinate system; the theoretical position information includes the center point coordinates of the part, the geometric features include the size of the part, the set of outer contour vertices, and the preset suction cup position, and the gripping mode information includes dual-head simultaneous gripping, dual-head dual gripping, or single-head single gripping; S22. Optimize the part picking order using a genetic algorithm: Initialize the population and randomly arrange the coordinates of the parts; Define fitness function

[0008] in, For adjacent grab points and The Euclidean distance between them This is a collision penalty item; S23. Select the population, perform crossover operation according to the preset crossover probability and mutation operation according to the preset mutation probability. Under the constraints of the part being located in the working area and the suction cup being aligned, iterate with the goal of minimizing the fitness function. Terminate when the maximum number of iterations is reached or the fitness change is less than the threshold, and output the coarse positioning coordinate sequence corresponding to the optimal grasping order to the host computer.

[0009] Furthermore, the specific steps of step S3 are as follows: S31. Establish a structured suction cup parameter database to store the unique ID, physical parameters, performance parameters, and physical installation parameters of each suction cup; the physical parameters include diameter, height, material, and hardness; the performance parameters include maximum negative pressure value, response time, and leakage rate; S32. Calculate the minimum bounding rectangle and convex hull features based on the geometric features of the part; the geometric features include the point cloud data of the part boundary; S33. Obtain several candidate suction cups from the suction cup parameter database. For each candidate suction cup, calculate the shortest Euclidean distance between the center point of the suction cup and the contour of the part, and apply the following distance constraint formula to filter and eliminate suction cups that do not meet the conditions, thereby obtaining a set of valid candidate suction cups:

[0010] in, The coordinates of the suction cup center are: For the part contour point set, The preset safe distance threshold; S34. The optimal suction cup combination is solved using a multi-objective genetic algorithm according to the following objective function and constraints; Objective function:

[0011] Where k is the number of suction cups in the suction cup assembly. Let be the distance between the i-th suction cup and the contour of the part. The total gripping force of the candidate suction cup combination is represented by w1 and w2, which are weighting coefficients. Constraints: Suction cup spacing ≥ preset spacing multiple multiplied by suction cup diameter, and total negative pressure value ≥ preset safety factor multiplied by part weight; S35. The optimal suction cup combination information is sent to the robotic arm controller. Before gripping, the actual negative pressure value is detected by the force sensor. When the deviation between the actual negative pressure value and the theoretical negative pressure value exceeds the preset deviation threshold, a recalculation is triggered, and the abnormal situation is recorded in the database for process optimization.

[0012] Furthermore, the specific steps of step S4 are as follows: S41. The host computer receives the coarse positioning coordinate sequence and the optimal suction cup combination information, and extracts key coordinate information for motion planning. The key coordinate information includes the center point coordinates of the part, whether it is necessary to open both suction cup arms to grasp at the same time, whether the suction cups need to rotate at an angle, and the area of ​​the part. S42. The host computer uses a path optimization algorithm to perform linear interpolation and circular interpolation calculations on the coordinates of adjacent gripping points in the coarse positioning coordinate sequence to generate a smooth joint space trajectory; the gripping point coordinates refer to the coarse positioning positions of each part that the robotic arm's end-effector carrying the suction cup needs to move to. S43. The joint space trajectory is encapsulated according to a preset communication protocol format, compiled to generate motion instructions containing target position, preset velocity curve type and acceleration limit parameters, and then transmitted to the robotic arm controller in real time via industrial bus. S44. The robotic arm performs coarse positioning, moves to the globally estimated position at a first speed ratio based on the center coordinates of the part, and then performs a spiral search path within a preset range around the target point, moving with a preset step accuracy until the vision inspection device successfully captures the preset marker point. S45. During the coarse positioning process, real-time monitoring of joint angles and feedback from external sensors is used for safe obstacle avoidance; If no marker is detected for a preset number of consecutive frames, a backtracking search strategy is initiated, and the scan is re-tracked along the original path for a preset distance. If the visual feedback deviation exceeds the preset positioning deviation threshold, the robotic arm is controlled to slow down to the preset second speed ratio and reposition itself.

[0013] Furthermore, the specific steps of step S5 are as follows: S51. Obtain the coordinates of the marker image and determine the local search area where the part is located based on the coordinates of the marker image; S52. Control the visual inspection device to acquire images of the local search area, use an image enhancement algorithm to enhance the contrast of the acquired images, and use an image fusion algorithm to stitch together multi-view images to obtain the target image to be processed. S53. Perform feature point detection and sub-pixel edge extraction on the target image, identify the feature points and boundary contours of the part, and combine the camera calibration parameters and ambient temperature compensation values ​​to calculate the precise three-dimensional coordinates of the part in the visual coordinate system and generate the precise coordinates of the part. S54. For the same target image, perform template matching, deep learning classification and physical size verification in sequence to determine whether there is a part at the target location, whether the type and specifications of the part are consistent with the expectation, and whether the part has a signal. S55. Execute collision avoidance warning based on whether the part has a signal and the spatial distance between the robotic arm and surrounding objects under real-time monitoring; When the distance between the robotic arm and surrounding objects is detected to be less than a preset warning threshold, a graded alarm is triggered and an alarm signal is output to control the robotic arm to decelerate or brake urgently.

[0014] Furthermore, the specific steps of step S52 are as follows: S521. Control the industrial camera in the vision inspection device to acquire images of the local search area in conjunction with the ring LED backlight, and dynamically adjust the exposure time according to the ambient light intensity; S522. Apply the multi-scale Retinex algorithm to the acquired raw images for image enhancement to improve imaging quality in low-light environments; S523. Based on the Poisson fusion algorithm, images acquired from multiple perspectives are stitched together to obtain a panoramic image covering the local search area, which is used as the target image after fusion. The specific steps of step S53 are as follows: S531. Receive the fused target image and use the Harris-Laplace corner detector to identify the feature points of the part; S532. Sub-pixel edge extraction is performed on the feature points identified by Zernike moments to achieve sub-pixel level localization; S533. Construct a two-dimensional Gaussian pyramid to perform multi-scale matching on the extracted features in order to eliminate noise interference; S534. The Tsai two-step method is used to calibrate the vision inspection device, and length compensation is performed in combination with real-time temperature sensor data. Then, the accurate three-dimensional coordinates of the part in the vision coordinate system are calculated based on the compensated calibration parameters, which are used as the accurate coordinates of the part. The specific steps of step S54 are as follows: S541. Receive the fused target image, perform template matching detection, and quickly determine whether there is a part in the target area; S542. Perform deep learning classification on the image regions that are matched by the template to identify the type of part and the state of its defects; S543. Perform physical size verification on image regions classified by deep learning, and measure the length, width and height of parts using OpenCV contour analysis to verify specification consistency; S544. Generate a signal indicating whether a part exists in the target area, the type and defect status of the part, and the length, width and height of the part. The specific steps of step S55 are as follows: S551. Receive whether the part has a signal and acquire the angle data of each joint of the robotic arm and the feedback from external sensors in real time; S552. An octree spatial segmentation algorithm is applied to monitor the spatial distance between the robotic arm and surrounding objects in real time, and a safety threshold area is set, which includes a warning area above the part and a danger area below it. S553. When the detected distance is less than the preset warning threshold, a tiered alarm is triggered based on the distance value: A yellow alert signal is output when the distance is less than the first threshold. A red alarm signal is output when the distance is less than the second threshold. An emergency braking signal is output when the distance is less than the third threshold. S554. Output the alarm signal to the three-color light for status indication, and at the same time send the braking signal to the robotic arm controller to perform emergency braking to protect the suction cup.

[0015] Furthermore, the specific steps of step S6 are as follows: S61. Obtain the transformation matrix between the visual coordinate system and the robot arm base coordinate system; S62. Obtain the precise coordinates of the part in the visual coordinate system; S63. Obtain the base coordinate values ​​of the robotic arm based on vision measurement, and compare them with the theoretical position information of the part in the laser cutting machine coordinate system extracted in step S21; If the deviation between the two is within the preset range, the measured value is used; otherwise, an exception is triggered, and the precise coordinates of the gripping points of each part in the robot arm's base coordinate system are finally obtained.

[0016] Furthermore, the specific steps of step S7 are as follows: S71. Obtain the precise coordinates of the gripping point and the gripping mode information of the part; S72. In the case of a dual-head, dual-pickup mode, calculate the feeding coordinates of the two gripping axes using the following formula: A-axis unloading coordinate X = X-axis offset + part center point coordinate X + (claw 1 center point coordinate X - claw 2 center point coordinate X); A-axis feeding coordinate Y = Y-axis offset + part center point coordinate Y + (claw 1 center point coordinate Y - claw 2 center point coordinate Y); Among them, the center point coordinates of the parts are the JSON data parsed in step S2, and the center point coordinates of claw 1 and claw 2 are the physical installation parameters of the suction cup; S73. Based on the calculated coordinates of the gripping and unloading points, and combined with the gripping mode, generate the final unloading motion command to control the robotic arm to complete the unloading action.

[0017] Secondly, embodiments of this application also provide an adaptive sorting and unloading system for a laser cutting machine, including a vision inspection device, a robotic arm, a memory, and a processor; The visual inspection device is installed on the sorting equipment to collect images; The robotic arm is equipped with a suction cup matrix at its end. The memory contains computer programs and a database containing parameters for each suction cup; The processor, connected to the vision inspection device, the robotic arm, and the memory, executes the computer program to implement the following functional modules: The calibration module performs visual calibration on the vision inspection device and establishes the transformation relationship between the visual coordinate system and the robot arm's base coordinate system. The analysis and optimization module reads and parses the GCode code of the cutting program, obtains the theoretical position information and geometric features of each part, and uses a genetic algorithm to optimize the gripping order of the parts to generate a coarse positioning coordinate sequence. The suction cup matching module matches and outputs the optimal suction cup combination required to grip each part based on the geometric features of the part from the suction cup parameter database in memory. The motion control module generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies preset marker points on the part through a vision inspection device to obtain the image coordinates of the marker points; The vision processing module acquires images of the target area through a vision inspection device, performs image processing to accurately identify the position of the part, and simultaneously detects whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence of the part signal; The coordinate calculation module calculates the precise coordinates of the gripping points of each part in the robot arm's base coordinate system based on the transformation relationship between the visual coordinate system and the robot arm's base coordinate system and the precise coordinates of the parts. The unloading control module controls the robotic arm to perform unloading actions based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part.

[0018] As can be seen from the above technical solutions, this application has the following advantages: The adaptive sorting and unloading method and system for laser cutting machines provided in this application obtains the prior geometric and positional information of parts by parsing GCode codes, and generates the optimal gripping sequence by combining genetic algorithms. This eliminates the blind search process in traditional vision sorting and reduces the length of the robotic arm's movement trajectory and the operation cycle time. Based on the mapping relationship between the structured suction cup parameter database and the geometric features of the parts, a multi-objective optimization algorithm is used to dynamically match the optimal suction cup combination, solving the problem that a single end effector cannot adapt to the sorting of multi-specification and irregularly shaped parts. A multi-level positioning strategy of theoretical coordinate coarse positioning, marker point guided search, and sub-pixel fine positioning is adopted, combined with coordinate system transformation matrix and temperature compensation algorithm, to eliminate mechanical repetitive positioning errors, thermal drift, and ambient light interference, thereby improving gripping accuracy. The integration of part existence detection and a real-time anti-collision warning mechanism based on octree spatial segmentation can identify the risk of empty gripping before execution and monitor spatial interference during movement to prevent equipment collisions and workpiece damage. By using force sensors to feedback the deviation between the actual negative pressure value and the theoretical value, a recalculation mechanism is triggered and abnormal data is recorded, realizing real-time correction and long-term iterative optimization of the gripping process. Attached Figure Description

[0019] To more clearly illustrate the technical solution of this application, the accompanying drawings used in the description will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic flowchart of the adaptive sorting and unloading method for laser cutting machines according to the present invention.

[0021] Figure 2 This is a schematic diagram of the adaptive sorting and unloading system for laser cutting machines according to the present invention. Detailed Implementation

[0022] The various embodiments of this disclosure will be described more fully in the following detailed description of the specific steps of the adaptive sorting and unloading method for laser cutting machines. This disclosure may have various embodiments, and adjustments and changes may be made therein. However, it should be understood that there is no intention to limit the various embodiments of this disclosure to the specific embodiments disclosed herein, but rather this disclosure should be understood to cover all adjustments, equivalents, and / or alternatives falling within the spirit and scope of the various embodiments of this disclosure.

[0023] This embodiment provides an adaptive sorting and unloading method for laser cutting machines, which optimizes the gripping order by parsing GCode, uses multi-level visual positioning, and dynamically matches suction cups to achieve efficient and safe unloading.

[0024] 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.

[0025] Please see Figure 1 The diagram shows a flowchart of an adaptive sorting and unloading method for a laser cutting machine in a specific embodiment. The method includes the following steps: S1. Install a vision inspection device on the sorting equipment and perform vision calibration to establish the transformation relationship between the vision coordinate system and the robot arm base coordinate system; It should be noted that this step solves the data mapping problem between heterogeneous coordinate systems, providing a foundation for visual perception data to drive the movement of the robotic arm and eliminating positioning deviations caused by inconsistencies in coordinate systems. S2. Read and parse the GCode of the cutting program to obtain the theoretical position information and geometric features of each part, and use a genetic algorithm to optimize the gripping order of the parts and generate a coarse positioning coordinate sequence; It should be noted that the steps use processing source data to reduce the reliance on pure visual search, and minimize the idle travel time and collision probability of the robotic arm through a global path optimization algorithm, thereby improving the overall efficiency of the sorting operation. S3. Establish a database containing parameters of each suction cup, and based on the geometric features of the parts, match and output the optimal suction cup combination required to grip each part from the database; It should be noted that this step enables dynamic configuration of the end effector, ensuring that the suction cup layout matches the part's contour and weight distribution, thus avoiding gripping failure or workpiece deformation due to improper actuator selection. S4. The host computer generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies the preset marker points on the part through the vision inspection device to obtain the image coordinates of the marker points; It should be noted that in this step, the prior coordinates guide the robotic arm to quickly approach the target area, and the identification of marker points narrows the large-scale search to a local area, which shortens the positioning time and improves the targeting of the search. S5. Acquire images of the target area through a vision inspection device, perform image processing to accurately identify the position of the part, and simultaneously detect whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence signal of the part; It should be noted that this step achieves high-precision local feature extraction, eliminating residual errors in macroscopic positioning; at the same time, through the existence detection logic, it prevents malfunctions in the event of missing parts, thus protecting the equipment and tooling. S6. Based on the transformation relationship between the visual coordinate system and the robot arm base coordinate system and the precise coordinates of the parts, calculate the precise coordinates of the gripping points of each part in the robot arm base coordinate system; It should be noted that this step completes the final calculation from image pixel coordinates to executable coordinates in robot joint space, ensuring that the end effector of the robotic arm can accurately reach the preset grasping pose. S7. Based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part, control the robotic arm to perform the unloading action; It should be noted that this step executes the final task based on the generated optimized path and precise coordinates, realizing an automated closed loop in the sorting process and ensuring the accuracy, stability, and efficiency of the material unloading action.

[0026] This embodiment uses genetic algorithms to optimize the grasping sequence by parsing GCode, and combines visual multi-level positioning and coordinate calibration to eliminate accumulated errors; based on geometric features, it dynamically matches the optimal suction cup combination, integrates part presence detection and spatial anti-collision warning, and realizes high-precision, high-efficiency and high-safety adaptive sorting and unloading of laser-cut parts.

[0027] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to fully illustrate the specific implementation process in this embodiment, another adaptive sorting and unloading method for laser cutting machines is provided, which includes the following steps: S1. Install a vision inspection device on the sorting equipment and perform vision calibration to establish the transformation relationship between the vision coordinate system and the robot arm base coordinate system; the specific steps of step S1 are as follows: S11. Based on the gripping motion direction of the sorting equipment, a vision inspection device is installed on a suction cup bracket. The vision inspection device includes a camera, a ring-shaped infrared LED light source, and a diffuser plate. S12. Using the front crossbeam frame and the left crossbeam frame of the laser cutting machine table as a reference, adjust the installation height and shooting angle of the vision inspection device through the calibration tool so that the field of view of the vision inspection device covers the entire laser cutting machine table. S13. Adjust the illumination angle of the infrared light source so that the optical axis of the infrared light source deflects at a preset angle away from the baseline perpendicular to the laser cutting machine table, and the light is evenly diffused by the diffuser plate. S14. Obtain camera parameters and installation area range, calculate the theoretically recommended installation parameters required to cover the target monitoring area, compare the actual installation parameters with the theoretically recommended installation parameters, complete the installation compliance verification, and finally establish the transformation matrix between the visual coordinate system and the robotic arm base coordinate system; Through the above calibration steps, a transformation matrix from the visual coordinate system to the robot arm base coordinate system can be obtained. This matrix will be used in step S6 to transform the precise coordinates of the visually recognized part to the robot arm base coordinate system. S2. Read and parse the GCode of the cutting program to obtain the theoretical position information and geometric features of each part, and use a genetic algorithm to optimize the gripping order of the parts and generate a coarse positioning coordinate sequence; This step uses the prior information contained in the cutting program itself, avoiding a global visual search of all parts and reducing the processing burden on the vision system; at the same time, by optimizing the grasping order through a genetic algorithm, the idle travel time and motion path length of the robotic arm can be reduced. The specific steps of step S2 are as follows: S21. Load the GCode file, open and read the file content using the ifstream class, parse the string stream into a JSON object using the nlohmann library, access the fields of the JSON object through key-value pairs, and extract the theoretical position information, geometric features, and gripping mode information of each part in the laser cutting machine coordinate system; the theoretical position information includes the center point coordinates of the part, the geometric features include the size of the part, the set of outer contour vertices, and the preset suction cup position, and the gripping mode information includes dual-head simultaneous gripping, dual-head dual gripping, or single-head single gripping; Dual-head simultaneous picking is a robotic arm with two gripping heads at the end, which simultaneously grip the same large part (such as a long strip-shaped workpiece, which requires two points of support). Dual-head dual-inspection means that a robotic arm has two gripping heads at its end, which simultaneously grip multiple parts in the same stockpile (equivalent to gripping multiple parts at once); single-head single-pickup means that a robotic arm has only one gripping head (suction cup assembly) at its end, which grips one part at a time. S22. Optimize the part picking order using a genetic algorithm: Initialize the population and randomly arrange the coordinates of the parts; Define fitness function

[0028] in, For adjacent grab points and The Euclidean distance between them This is a collision penalty item; S23. Select the population, perform crossover operation according to the preset crossover probability (e.g., 0.7) and mutation operation according to the preset mutation probability (e.g., 0.01). Under the constraints of the part being located in the working area and the suction cup being aligned, iterate with the goal of minimizing the fitness function. Terminate when the maximum number of iterations is reached or the fitness change is less than the threshold, and output the coarse positioning coordinate sequence corresponding to the optimal gripping order to the host computer. The optimized coarse positioning coordinate sequence will serve as the basis for the robotic arm motion planning in step S4. At the same time, the geometric features of each part will be passed to step S3 for suction cup matching, and the gripping mode information will be passed to step S7 for material unloading control. S3. Establish a database containing parameters of each suction cup, and based on the geometric features of the parts, match and output the optimal suction cup combination required to grip each part from the database; This step enables dynamic configuration of the end effector, solving the problem that traditional fixed suction cup matrices cannot adapt to the sorting of parts of various specifications. By establishing a structured suction cup database and combining it with the geometric features of the parts for multi-objective optimization, the most suitable suction cup combination can be selected for each part, ensuring gripping stability and avoiding workpiece deformation. The specific steps of step S3 are as follows: S31. Establish a structured suction cup parameter database to store the unique ID, physical parameters, performance parameters, and physical installation parameters of each suction cup; the physical parameters include diameter, height, material, and hardness; the performance parameters include maximum negative pressure value, response time, and leakage rate; The suction cups in the suction cup parameter database are arranged in rows and columns to form a suction cup matrix, and each suction cup has a unique row and column number; the optimal suction cup combination obtained in the subsequent step S34 refers to a number of suction cups selected from the suction cup matrix and their activation states; S32. Calculate the minimum bounding rectangle and convex hull features based on the geometric features of the part; the geometric features include the point cloud data of the part boundary; S33. Obtain several candidate suction cups from the suction cup parameter database. For each candidate suction cup, calculate the shortest Euclidean distance between the center point of the suction cup and the contour of the part, and apply the following distance constraint formula to filter and eliminate suction cups that do not meet the conditions, thereby obtaining a set of valid candidate suction cups:

[0029] in, The coordinates of the suction cup center are: For the part contour point set, Set a preset safety distance threshold (e.g., 10mm) to ensure that the suction cup is completely within the projection range of the part; S34. The optimal suction cup combination is solved using a multi-objective genetic algorithm according to the following objective function and constraints; Objective function:

[0030] Where k is the number of suction cups in the suction cup assembly. Let be the distance between the i-th suction cup and the contour of the part. The total gripping force of the candidate suction cup combination is represented by w1 and w2, which are weighting coefficients. Constraints: The suction cup spacing must be greater than or equal to the preset spacing multiple (e.g., 2 times) multiplied by the suction cup diameter, and the total negative pressure value must be greater than or equal to the preset safety factor (e.g., 1.5 times) multiplied by the weight of the part. S35. Send the solved optimal suction cup combination information to the robotic arm controller, and detect the actual negative pressure value through the force sensor before grasping. When the deviation between the actual negative pressure value and the theoretical negative pressure value exceeds the preset deviation threshold, a recalculation is triggered, and the abnormal situation is recorded in the database for process optimization. The optimal suction cup combination information will serve as the basis for configuring the end effector of the robotic arm in step S4, ensuring that the correct suction cup combination has been activated when the robotic arm moves to the target position; S4. The host computer generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies the preset marker points on the part through the vision inspection device to obtain the image coordinates of the marker points; This step realizes the transformation from theoretical coordinates to actual motion. A smooth motion trajectory is generated through path optimization algorithm, and combined with a multi-stage coarse positioning strategy, the robotic arm can quickly and accurately approach the target position, providing good initial conditions for subsequent visual fine positioning; The specific steps of step S4 are as follows: S41. The host computer receives the coarse positioning coordinate sequence and the optimal suction cup combination information, and extracts key coordinate information for motion planning. The key coordinate information includes the center point coordinates of the part, whether it is necessary to open both suction cup arms to grasp at the same time, whether the suction cups need to rotate at an angle, and the area of ​​the part. S42. The host computer uses a path optimization algorithm to perform linear interpolation and circular interpolation calculations on the coordinates of adjacent gripping points in the coarse positioning coordinate sequence to generate a smooth joint space trajectory; the gripping point coordinates refer to the coarse positioning positions of each part that the robotic arm's end-effector carrying the suction cup needs to move to. S43. The joint space trajectory is encapsulated according to a preset communication protocol format (such as JSON format), compiled to generate motion instructions containing target position, preset velocity curve type (such as S-curve acceleration and deceleration) and acceleration limit parameters, and then transmitted to the robotic arm controller in real time via industrial bus (such as EtherCAT). S44. The robotic arm performs coarse positioning, moving to the globally estimated position at a first speed ratio (e.g., 70% of the maximum speed) based on the center coordinates of the part, and then performing a spiral search path within a preset range (e.g., ±50mm) around the target point, moving with a preset step accuracy (e.g., 2mm) until the vision inspection device successfully captures the preset marker point. S45. During the coarse positioning process, real-time monitoring of joint angles and feedback from external sensors is used for safe obstacle avoidance; If no marker is detected for a preset number of consecutive frames (e.g., 3 frames), a backtracking search strategy is initiated, and the original path is backtracked by a preset distance (e.g., 10mm) before rescanning. If the visual feedback deviation exceeds the preset positioning deviation threshold (e.g., ±2mm), the robotic arm is controlled to slow down to the preset second speed ratio (e.g., 30% of the maximum speed) and repositioned. The coordinates of the marker point image captured in step S44 will be used as the basis for determining the local search area in step S5, realizing the transition from global coarse localization to local fine localization. S5. Acquire images of the target area through a vision inspection device, perform image processing to accurately identify the position of the part, and simultaneously detect whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence signal of the part; This step is the core of visual precision positioning; by using marker points to guide and narrow the search range, the efficiency and accuracy of image processing are greatly improved; at the same time, it integrates the function of detecting the presence or absence of parts, which can identify the risk of empty grasp before grasping and protect the suction cup; the following is a detailed implementation of each sub-step; The specific steps of step S5 are as follows: S51. Obtain the coordinates of the marker image and determine the local search area where the part is located based on the coordinates of the marker image; S52. Control the visual inspection device to acquire images of the local search area, use an image enhancement algorithm to enhance the contrast of the acquired images, and use an image fusion algorithm to stitch together multi-view images to obtain the target image to be processed. The specific steps of step S52 are as follows: S521. Control the industrial camera in the vision inspection device to acquire images of the local search area in conjunction with the ring LED backlight, and dynamically adjust the exposure time according to the ambient light intensity (e.g., adjust in real time within the range of 1 / 1000s to 1 / 50s). S522. Apply the multi-scale Retinex algorithm to the acquired raw images for image enhancement to improve imaging quality in low-light environments; The multi-scale Retinex algorithm is an image enhancement algorithm based on retinal cortex theory. It improves imaging quality in low-light environments by estimating illumination and reflection components at different scales. S523. Based on the Poisson fusion algorithm, images acquired from multiple perspectives are stitched together to obtain a panoramic image covering the local search area, which is used as the target image after fusion. S53. Perform feature point detection and sub-pixel edge extraction on the target image to identify the feature points and boundary contours of the part. Combine camera calibration parameters and ambient temperature compensation values ​​to calculate the precise three-dimensional coordinates of the part in the visual coordinate system, generating the precise coordinates of the part. The specific steps of step S53 are as follows: S531. Receive the fused target image and use the Harris-Laplace corner detector to identify the feature points of the part; S532. Sub-pixel edge extraction is performed on the feature points identified by Zernike moments to achieve sub-pixel-level positioning with a positioning accuracy of ±0.01mm; Zernike moments are image moments based on orthogonal polynomials, which are rotationally invariant and can be used for precise edge localization at the subpixel level. S533. Construct a two-dimensional Gaussian pyramid to perform multi-scale matching on the extracted features in order to eliminate noise interference; Gaussian pyramids are a multi-scale image representation method that constructs image layers of different resolutions by performing Gaussian smoothing and downsampling on the image, which are then used for multi-scale feature matching. S534. The vision inspection device is calibrated using the Tsai two-step method, and length compensation is performed by combining real-time temperature sensor data (e.g., 0.002 mm compensation per °C temperature difference). The precise three-dimensional coordinates of the part in the vision coordinate system are then calculated based on the compensated calibration parameters, which are used as the precise coordinates of the part. Tsai's two-step method is a commonly used camera calibration method. It solves for the initial values ​​of camera parameters linearly and then performs nonlinear optimization to establish the mapping relationship between image pixel coordinates and spatial three-dimensional coordinates. S54. For the same target image, perform template matching, deep learning classification, and physical size verification sequentially to determine whether a part exists at the target location, whether the type and specifications of the part are consistent with expectations, and generate a signal indicating whether the part exists. The specific steps of step S54 are as follows: S541. Receive the fused target image, perform template matching detection, and quickly determine whether there is a part in the target area; S542. Perform deep learning classification on the image regions that are matched by the template to identify the type of part and the state of its defects; S543. Perform physical size verification on image regions classified by deep learning, and measure the length, width and height of parts using OpenCV contour analysis to verify specification consistency; S544. Generate a signal indicating whether a part exists in the target area, the type and defect status of the part, and the length, width and height of the part. S55. Based on whether the part has a signal and the real-time monitored spatial distance between the robotic arm and surrounding objects, execute a collision avoidance warning; the specific steps of step S55 are as follows: S551. Receive whether the part has a signal and acquire the angle data of each joint of the robotic arm and the feedback from external sensors in real time; S552. An octree spatial segmentation algorithm is applied to monitor the spatial distance between the robotic arm and surrounding objects in real time, and a safety threshold area is set, which includes a warning area (e.g., 10mm) above the part and a danger area (e.g., 5mm) below it. Octree spatial partitioning algorithm is a tree data structure that recursively divides three-dimensional space into eight subspaces, used to quickly detect the spatial distance between a robotic arm and surrounding objects; S553. When the detected distance is less than the preset warning threshold, a tiered alarm is triggered based on the distance value: A yellow warning signal is output when the distance is less than a first threshold (e.g., 30mm); A red alarm signal is output when the distance is less than the second threshold (e.g., 15mm); An emergency braking signal is output when the distance is less than the third threshold (e.g., 5mm); S554. Output the alarm signal to the three-color light for status indication, and at the same time send the braking signal to the robotic arm controller to perform emergency braking to protect the suction cup. When the distance between the robotic arm and surrounding objects is detected to be less than a preset warning threshold, a graded alarm is triggered and an alarm signal is output to control the robotic arm to decelerate or brake urgently. The precise coordinates of the part obtained in step S53 will be used as the input for step S6, and the presence or absence of the part signal obtained in step S54 will be used for the collision warning in step S55, forming a complete visual processing closed loop. S6. Based on the transformation relationship between the visual coordinate system and the robot arm base coordinate system and the precise coordinates of the parts, calculate the precise coordinates of the gripping points of each part in the robot arm base coordinate system; This step completes the final calculation from image pixel coordinates to executable coordinates in the robot joint space. Using the transformation matrix established in step S1, the precisely recognized coordinates are transformed to the robot arm's base coordinate system and then superimposed and calibrated with the theoretical position information to eliminate accumulated errors. The specific steps of step S6 are as follows: S61. Obtain the transformation matrix between the visual coordinate system and the robot arm base coordinate system; S62. Obtain the precise coordinates of the part in the visual coordinate system; S63. Obtain the robot arm base coordinate values ​​based on vision measurement and compare them with the theoretical position information (JSON data) of the part in the laser cutting machine coordinate system extracted in step S21; If the deviation between the two is within the preset range, the measured value is used; otherwise, an exception is triggered, and the precise coordinates of the gripping points of each part in the robot arm's base coordinate system are finally obtained. The precise coordinates of the calibrated gripping point will serve as the final target position for the unloading action in step S7, ensuring that the end effector of the robotic arm can accurately reach the location of the part. S7. Based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part, control the robotic arm to perform the unloading action; This step is the final execution stage of the method, completing the unloading operation based on the generated optimized path and precise coordinates; for special gripping modes such as dual-head dual-pickup, coordinate conversion is required based on the physical installation parameters of the suction cup to ensure that the two gripping axes can work together. The specific steps of step S7 are as follows: S71. Obtain the precise coordinates of the gripping point and the gripping mode information of the part; S72. In the case of a dual-head, dual-pickup mode, calculate the feeding coordinates of the two gripping axes using the following formula: A-axis unloading coordinate X = X-axis offset + part center point coordinate X + (claw 1 center point coordinate X - claw 2 center point coordinate X); A-axis feeding coordinate Y = Y-axis offset + part center point coordinate Y + (claw 1 center point coordinate Y - claw 2 center point coordinate Y); Among them, the center point coordinates of the parts are the JSON data parsed in step S2, and the center point coordinates of claw 1 and claw 2 are the physical installation parameters of the suction cup; S73. Based on the calculated coordinates of the gripping and unloading points, and combined with the gripping mode, generate the final unloading motion command to control the robotic arm to complete the unloading action.

[0031] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0032] like Figure 2 As shown, the following are embodiments of the adaptive sorting and unloading system for laser cutting machines provided in this disclosure. This system and the adaptive sorting and unloading methods for laser cutting machines in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the adaptive sorting and unloading system for laser cutting machines, please refer to the embodiments of the adaptive sorting and unloading methods for laser cutting machines described above.

[0033] The system includes a vision inspection device, a robotic arm, a memory, and a processor; The visual inspection device is installed on the sorting equipment to collect images; The end effector of the robotic arm is equipped with a suction cup matrix, which contains multiple vacuum suction cups arranged in rows and columns. The memory contains computer programs and a database containing parameters for each suction cup; The processor, connected to the vision inspection device, the robotic arm, and the memory, executes the computer program to implement the following functional modules: The calibration module performs visual calibration on the vision inspection device and establishes the transformation relationship between the visual coordinate system and the robot arm's base coordinate system. The analysis and optimization module reads and parses the GCode code of the cutting program, obtains the theoretical position information and geometric features of each part, and uses a genetic algorithm to optimize the gripping order of the parts to generate a coarse positioning coordinate sequence. The suction cup matching module matches and outputs the optimal suction cup combination required to grip each part based on the geometric features of the part from the suction cup parameter database in memory. The motion control module generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies preset marker points on the part through a vision inspection device to obtain the image coordinates of the marker points; The vision processing module acquires images of the target area through a vision inspection device, performs image processing to accurately identify the position of the part, and simultaneously detects whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence of the part signal; The coordinate calculation module calculates the precise coordinates of the gripping points of each part in the robot arm's base coordinate system based on the transformation relationship between the visual coordinate system and the robot arm's base coordinate system and the precise coordinates of the parts. The unloading control module controls the robotic arm to perform unloading actions based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part.

[0034] This embodiment achieves high-precision adaptive sorting through the interactive collaboration of the calibration module, analysis and optimization module, suction cup matching module, motion control module, vision processing module, coordinate calculation module, and unloading control module. Based on GCode parsing and genetic algorithm path optimization, it adopts a combination of coarse and fine vision positioning and temperature compensation to dynamically match the optimal suction cup and integrate multi-level safety detection.

[0035] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An adaptive sorting and unloading method for a laser cutting machine, characterized in that, Includes the following steps: S1. Install a vision inspection device on the sorting equipment and perform vision calibration to establish the transformation relationship between the vision coordinate system and the robot arm base coordinate system; S2. Read and parse the GCode of the cutting program to obtain the theoretical position information and geometric features of each part, and use a genetic algorithm to optimize the gripping order of the parts and generate a coarse positioning coordinate sequence; S3. Establish a database containing parameters of each suction cup, and based on the geometric features of the parts, match and output the optimal suction cup combination required to grip each part from the database; S4. The host computer generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies the preset marker points on the part through the vision inspection device to obtain the image coordinates of the marker points; S5. Determine the search area based on the coordinates of the marker image, acquire the image of the search area through the vision inspection device, perform image processing to accurately identify the position of the part, and simultaneously detect whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence of the part signal; S6. Based on the transformation relationship between the visual coordinate system and the robot arm base coordinate system and the precise coordinates of the parts, calculate the precise coordinates of the gripping points of each part in the robot arm base coordinate system; S7. Based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part, control the robotic arm to perform the unloading action.

2. The adaptive sorting and unloading method for a laser cutting machine according to claim 1, characterized in that, The specific steps of step S1 are as follows: S11. Based on the gripping motion direction of the sorting equipment, a vision inspection device is installed on a suction cup bracket. The vision inspection device includes a camera, a ring-shaped infrared LED light source, and a diffuser plate. S12. Using the front crossbeam frame and the left crossbeam frame of the laser cutting machine table as a reference, adjust the installation height and shooting angle of the vision inspection device through the calibration tool so that the field of view of the vision inspection device covers the entire laser cutting machine table. S13. Adjust the illumination angle of the infrared light source so that the optical axis of the infrared light source deflects at a preset angle away from the baseline perpendicular to the laser cutting machine table, and the light is evenly diffused by the diffuser plate. S14. Obtain camera parameters and installation area range, calculate the theoretically recommended installation parameters required to cover the target monitoring area, compare the actual installation parameters with the theoretically recommended installation parameters, complete the installation compliance verification, and finally establish the transformation matrix between the visual coordinate system and the robotic arm base coordinate system.

3. The adaptive sorting and unloading method for a laser cutting machine according to claim 1, characterized in that, The specific steps of step S2 are as follows: S21. Load the GCode file, open and read the file content using the ifstream class, parse the string stream into a JSON object using the nlohmann library, access the fields of the JSON object through key-value pairs, and extract the theoretical position information, geometric features, and gripping mode information of each part in the laser cutting machine coordinate system; the theoretical position information includes the center point coordinates of the part, the geometric features include the size of the part, the set of outer contour vertices, and the preset suction cup position, and the gripping mode information includes dual-head simultaneous gripping, dual-head dual gripping, or single-head single gripping; S22. Optimize the part picking order using a genetic algorithm: Initialize the population and randomly arrange the coordinates of the parts; Define fitness function in, For adjacent grab points and The Euclidean distance between them This is a collision penalty item; S23. Select the population, perform crossover operation according to the preset crossover probability and mutation operation according to the preset mutation probability. Under the constraints of the part being located in the working area and the suction cup being aligned, iterate with the goal of minimizing the fitness function. Terminate when the maximum number of iterations is reached or the fitness change is less than the threshold, and output the coarse positioning coordinate sequence corresponding to the optimal grasping order to the host computer.

4. The adaptive sorting and unloading method for a laser cutting machine according to claim 2, characterized in that, The specific steps of step S3 are as follows: S31. Establish a structured suction cup parameter database to store the unique ID, physical parameters, performance parameters, and physical installation parameters of each suction cup; the physical parameters include diameter, height, material, and hardness; the performance parameters include maximum negative pressure value, response time, and leakage rate; S32. Calculate the minimum bounding rectangle and convex hull features based on the geometric features of the part; the geometric features include the point cloud data of the part boundary; S33. Obtain several candidate suction cups from the suction cup parameter database. For each candidate suction cup, calculate the shortest Euclidean distance between the center point of the suction cup and the contour of the part, and apply the following distance constraint formula to filter and eliminate suction cups that do not meet the conditions, thereby obtaining a set of valid candidate suction cups: in, The coordinates of the suction cup center are: For the part contour point set, The preset safe distance threshold; S34. The optimal suction cup combination is solved using a multi-objective genetic algorithm according to the following objective function and constraints; Objective function: Where k is the number of suction cups in the suction cup assembly. Let be the distance between the i-th suction cup and the contour of the part. The total gripping force of the candidate suction cup combination is represented by w1 and w2, which are weighting coefficients. Constraints: Suction cup spacing ≥ preset spacing multiple multiplied by suction cup diameter, and total negative pressure value ≥ preset safety factor multiplied by part weight; S35. The optimal suction cup combination information is sent to the robotic arm controller. Before gripping, the actual negative pressure value is detected by the force sensor. When the deviation between the actual negative pressure value and the theoretical negative pressure value exceeds the preset deviation threshold, a recalculation is triggered, and the abnormal situation is recorded in the database for process optimization.

5. The adaptive sorting and unloading method for a laser cutting machine according to claim 1, characterized in that, The specific steps of step S4 are as follows: S41. The host computer receives the coarse positioning coordinate sequence and the optimal suction cup combination information, and extracts key coordinate information for motion planning. The key coordinate information includes the center point coordinates of the part, whether it is necessary to open both suction cup arms to grasp at the same time, whether the suction cups need to rotate at an angle, and the area of ​​the part. S42. The host computer uses a path optimization algorithm to perform linear interpolation and circular interpolation calculations on the coordinates of adjacent gripping points in the coarse positioning coordinate sequence to generate a smooth joint space trajectory; the gripping point coordinates refer to the coarse positioning positions of each part that the robotic arm's end-effector carrying the suction cup needs to move to. S43. The joint space trajectory is encapsulated according to a preset communication protocol format, compiled to generate motion instructions containing target position, preset velocity curve type and acceleration limit parameters, and then transmitted to the robotic arm controller in real time via industrial bus. S44. The robotic arm performs coarse positioning, moves to the globally estimated position at a first speed ratio based on the center coordinates of the part, and then performs a spiral search path within a preset range around the target point, moving with a preset step accuracy until the vision inspection device successfully captures the preset marker point. S45. During the coarse positioning process, real-time monitoring of joint angles and feedback from external sensors is used for safe obstacle avoidance. If no marker is detected for a consecutive preset number of frames, a backtracking search strategy is initiated, and the original path is backtracked for a preset distance before rescanning. If the visual feedback deviation exceeds the preset positioning deviation threshold, the robotic arm is controlled to slow down to the preset second speed ratio and reposition itself.

6. The adaptive sorting and unloading method for a laser cutting machine according to claim 4, characterized in that, The specific steps of step S5 are as follows: S51. Obtain the coordinates of the marker point image, and determine the local search area where the part is located based on the coordinates of the marker point image; S52. Control the visual inspection device to acquire images of the local search area, use an image enhancement algorithm to enhance the contrast of the acquired images, and use an image fusion algorithm to stitch together multi-view images to obtain the target image to be processed. S53. Perform feature point detection and sub-pixel edge extraction on the target image, identify the feature points and boundary contours of the part, and combine the camera calibration parameters and ambient temperature compensation values ​​to calculate the precise three-dimensional coordinates of the part in the visual coordinate system and generate the precise coordinates of the part. S54. For the same target image, perform template matching, deep learning classification and physical size verification in sequence to determine whether there is a part at the target location, whether the type and specifications of the part are consistent with the expectation, and whether the part has a signal. S55. Execute collision avoidance warning based on whether the part has a signal and the spatial distance between the robotic arm and surrounding objects under real-time monitoring; When the distance between the robotic arm and surrounding objects is detected to be less than a preset warning threshold, a graded alarm is triggered and an alarm signal is output to control the robotic arm to decelerate or brake urgently.

7. The adaptive sorting and unloading method for a laser cutting machine according to claim 6, characterized in that, The specific steps of step S52 are as follows: S521. Control the industrial camera in the vision inspection device to acquire images of the local search area in conjunction with the ring LED backlight, and dynamically adjust the exposure time according to the ambient light intensity; S522. Apply the multi-scale Retinex algorithm to the acquired raw images for image enhancement to improve imaging quality in low-light environments; S523. Based on the Poisson fusion algorithm, images acquired from multiple perspectives are stitched together to obtain a panoramic image covering the local search area, which is used as the target image after fusion. The specific steps of step S53 are as follows: S531. Receive the fused target image and use the Harris-Laplace corner detector to identify the feature points of the part; S532. Sub-pixel edge extraction is performed on the feature points identified by Zernike moments to achieve sub-pixel level localization; S533. Construct a two-dimensional Gaussian pyramid to perform multi-scale matching on the extracted features in order to eliminate noise interference; S534. The Tsai two-step method is used to calibrate the vision inspection device, and length compensation is performed in combination with real-time temperature sensor data. Then, the accurate three-dimensional coordinates of the part in the vision coordinate system are calculated based on the compensated calibration parameters, which are used as the accurate coordinates of the part. The specific steps of step S54 are as follows: S541. Receive the fused target image, perform template matching detection, and quickly determine whether there is a part in the target area; S542. Perform deep learning classification on the image regions that are matched by the template to identify the type of part and the state of its defects; S543. Perform physical size verification on image regions classified by deep learning, and measure the length, width and height of parts using OpenCV contour analysis to verify specification consistency; S544. Generate a signal indicating whether a part exists in the target area, the type and defect status of the part, and the length, width and height of the part. The specific steps of step S55 are as follows: S551. Receive whether the part has a signal and acquire the angle data of each joint of the robotic arm and the feedback from external sensors in real time; S552. An octree spatial segmentation algorithm is applied to monitor the spatial distance between the robotic arm and surrounding objects in real time, and a safety threshold area is set, which includes a warning area above the part and a danger area below it. S553. When the detected distance is less than the preset warning threshold, a tiered alarm is triggered based on the distance value: A yellow alert signal is output when the distance is less than the first threshold. A red alarm signal is output when the distance is less than the second threshold. An emergency braking signal is output when the distance is less than the third threshold. S554. Output the alarm signal to the three-color light for status indication, and at the same time send the braking signal to the robotic arm controller to perform emergency braking to protect the suction cup.

8. The adaptive sorting and unloading method for a laser cutting machine according to claim 6, characterized in that, The specific steps of step S6 are as follows: S61. Obtain the transformation matrix between the visual coordinate system and the robot arm base coordinate system; S62. Obtain the precise coordinates of the part in the visual coordinate system; S63. Obtain the base coordinate values ​​of the robotic arm based on vision measurement, and compare them with the theoretical position information of the part in the laser cutting machine coordinate system extracted in step S21; If the deviation between the two is within the preset range, the measured value is used; otherwise, an exception is triggered, and the precise coordinates of the gripping points of each part in the robot arm's base coordinate system are finally obtained.

9. The adaptive sorting and unloading method for a laser cutting machine according to claim 6, characterized in that, The specific steps of step S7 are as follows: S71. Obtain the precise coordinates of the gripping point and the gripping mode information of the part; S72. In the case of a dual-head, dual-pickup mode, calculate the feeding coordinates of the two gripping axes using the following formula: A-axis unloading coordinate X = X-axis offset + part center point coordinate X + (claw 1 center point coordinate X - claw 2 center point coordinate X); A-axis feeding coordinate Y = Y-axis offset + part center point coordinate Y + (claw 1 center point coordinate Y - claw 2 center point coordinate Y); Among them, the center point coordinates of the parts are the JSON data parsed in step S2, and the center point coordinates of claw 1 and claw 2 are the physical installation parameters of the suction cup; S73. Based on the calculated coordinates of the gripping and unloading points, and combined with the gripping mode, generate the final unloading motion command to control the robotic arm to complete the unloading action.

10. An adaptive sorting and unloading system for a laser cutting machine, characterized in that, Includes visual inspection devices, robotic arms, memory, and processors; The visual inspection device is installed on the sorting equipment to collect images; The robotic arm is equipped with a suction cup matrix at its end. The memory contains the computer program and a database containing parameters for each suction cup; The processor, connected to the vision inspection device, the robotic arm, and the memory, executes the computer program to implement the following functional modules: The calibration module performs visual calibration on the vision inspection device and establishes the transformation relationship between the visual coordinate system and the robot arm's base coordinate system. The analysis and optimization module reads and parses the GCode code of the cutting program, obtains the theoretical position information and geometric features of each part, and uses a genetic algorithm to optimize the gripping order of the parts to generate a coarse positioning coordinate sequence. The suction cup matching module matches and outputs the optimal suction cup combination required to grip each part based on the geometric features of the part from the suction cup parameter database in the memory. The motion control module generates motion commands based on the coarse positioning coordinate sequence, controls the robotic arm to move to the target position with the matching optimal suction cup combination, and identifies preset marker points on the part through a vision inspection device to obtain the image coordinates of the marker points; The vision processing module determines the search area based on the coordinates of the marker point image, acquires the image of the search area through the vision inspection device, performs image processing to accurately identify the position of the part, and simultaneously detects whether the part exists at the target position, outputting the precise coordinates of the part and the presence or absence of the part signal. The coordinate calculation module calculates the precise coordinates of the gripping points of each part in the robot arm's base coordinate system based on the transformation relationship between the visual coordinate system and the robot arm's base coordinate system and the precise coordinates of the parts. The unloading control module controls the robotic arm to perform unloading actions based on the precise coordinates of the gripping point and the theoretical position information and geometric features of each part.

Images