A robotic arm positioning and picking method and system for ultrasonic welding of a medicator

Abstract

A method and system for robotic arm positioning and picking of ultrasonic welding of a drug dispenser, which relates to the technical field of robotic arm positioning and picking, obtains image data of the drug dispenser and preprocesses it, extracts features based on the preprocessed image data of the drug dispenser, performs three-dimensional positioning and pose analysis on the drug dispenser based on the extracted features, and plans the robotic arm clamping path based on the obtained pose analysis result and three-dimensional positioning result. Through image data preprocessing and pose analysis, the present invention realizes three-dimensional positioning of the drug dispenser, optimizes the clamping and welding paths of the robotic arm through path planning, ensures the accuracy of the welding process, and performs real-time detection and feedback on the welding quality through an image monitoring system, effectively improving the welding quality and production efficiency, enhancing the automation level and operation accuracy of the overall method, and significantly reducing manual intervention and the defective product rate.

Description

Technical Field

[0001] The present invention relates to the technical field of robotic arm positioning and picking, and particularly to a robotic arm positioning and picking method and system for ultrasonic welding of a medicator. Background Art

[0002] With the continuous development of the medical device industry, as an auxiliary medical device, the medicator has increasingly high requirements for the accuracy of its assembly and welding during the production process. The existing medicator assembly and welding processes usually adopt ultrasonic welding technology to ensure the stability of the welding part. In a complex production line, how to effectively improve the assembly accuracy and welding efficiency, especially the need for positioning, picking and placing, and welding with the help of automated equipment, is becoming increasingly urgent. The application of robotic arms in production has been gradually widely promoted, and through its flexible operating performance, automated operation can be achieved.

[0003] Currently, during the automated welding process of the medicator, robotic arms usually rely on sensors and image processing technologies for precise positioning and control. However, there are many deficiencies in the existing technologies;

[0004] First of all, during the transportation and positioning of the medicator, due to inaccurate or delayed data obtained by sensors, the robotic arm is prone to deviation when performing positioning and picking and placing operations;

[0005] Secondly, the real-time monitoring during the welding process is also relatively weak, and the welding quality cannot be comprehensively grasped, which may cause problems such as insecure product welding or uneven weld seams;

[0006] Moreover, the lack of effective path planning and motion control technologies makes it difficult to balance the speed and accuracy of the robotic arm during clamping, transferring, and welding operations.

[0007] Therefore, it is urgent to propose a robotic arm positioning and picking method and system for ultrasonic welding of a medicator. Summary of the Invention

[0008] The present invention provides a robotic arm positioning and picking method and system for ultrasonic welding of a medicator, aiming to solve the practical problem that the robotic arm is prone to deviation when performing positioning and picking and placing operations in related technologies, avoid the technical problem of insecure product welding caused by the lack of real-time monitoring during welding, and avoid the subsequent problem of poor accuracy of the robotic arm due to the lack of effective path planning and motion control technologies.

[0009] To achieve the above object, the embodiments of the present application are implemented as follows:

[0010] In a first aspect, the embodiments of the present application provide a robotic arm positioning and picking method for ultrasonic welding of a medicator, which is characterized by including the following:

[0011] Step S100: Based on the conveyor belt to transport the applicator, obtain the image data of the applicator and preprocess it;

[0012] Step S200: Extract features based on the preprocessed image data of the applicator, and perform three-dimensional positioning and attitude analysis on the applicator based on the extracted features;

[0013] Step S300: Plan the gripping path of the robotic arm based on the obtained attitude analysis result and three-dimensional positioning result;

[0014] Step S400: Based on the planned gripping path, the robotic arm grips the applicator and transports it to the mechanical turntable for ultrasonic welding, and monitors the image of the applicator during the welding process;

[0015] Step S500: Plan the path of the robotic arm at the other end based on the image data of the welded applicator;

[0016] Step S600: The robotic arm at the other end grips and transfers the processed applicator to the conveyor belt based on the planned path, and the conveyor belt transports the applicator at a constant speed.

[0017] Based on the conveyor belt to transport the applicator, obtain the image data of the applicator and preprocess it, extract features based on the preprocessed image data of the applicator, perform three-dimensional positioning and attitude analysis on the applicator based on the extracted features, plan the gripping path of the robotic arm based on the obtained attitude analysis result and three-dimensional positioning result, based on the planned gripping path, the robotic arm grips the applicator and transports it to the mechanical turntable for ultrasonic welding, and monitors the image of the applicator during the welding process, plan the path of the robotic arm at the other end based on the image data of the welded applicator, and the robotic arm at the other end grips and transfers the processed applicator to the conveyor belt based on the planned path, and the conveyor belt transports the applicator at a constant speed.

[0018] Combined with the first aspect, in some specific embodiments, based on the conveyor belt to transport the applicator, obtain the image data of the applicator and preprocess it, specifically including:

[0019] Step S101: Obtaining the image data of the applicator in the conveying state includes feeding processing and shooting processing;

[0020] Step S101.1: Feeding processing, manually placing the pre-assembled applicator on the conveyor belt, and the conveyor belt transports it at a constant speed;

[0021] Step S101.2: Obtain the image data of the applicator in the conveying state, by presetting a shooting area above the conveying area of the applicator, installing an industrial camera above the shooting area, and shooting the applicator to obtain continuous image data;

[0022] The shooting frequency of the camera is adjusted according to the speed of the conveyor belt and the size of the dispenser to ensure that a clear and complete image of the dispenser is captured as the conveyor belt transports it to the designated shooting area;

[0023] Step S102, preprocessing of the dispenser image data. The preprocessing includes image denoising, grayscale processing, edge detection, and image enhancement;

[0024] Step S102.1, image denoising. A Gaussian filter combined with median filtering is used to smooth the image, and the denoised dispenser image data is obtained;

[0025] Step S102.2, grayscale processing. The color image after denoising is converted into a grayscale image. The weighted average method is used to perform grayscale processing with the weights of the three channels, and the grayscale processed dispenser image data is obtained;

[0026] Step S102.3, edge detection. The Canny edge detection algorithm is used to extract the edges of the grayscale processed image, highlighting the contour of the dispenser image data;

[0027] Step S102.4, image enhancement. Histogram equalization is used to enhance the contrast of the image, enhancing the contour of the dispenser image data;

[0028] Step S102.5, based on steps S102.1 to S102.4, the preprocessed dispenser image data is obtained.

[0029] Combined with the first aspect, in some specific embodiments, feature extraction is performed based on the preprocessed dispenser image data, and three-dimensional positioning and pose analysis of the dispenser are performed based on the extracted features. Specifically, it includes:

[0030] Step S201, feature extraction. Based on the detection results obtained in step S102.3, the complete contour information in the dispenser image data is extracted, and the positions of the edge feature points in the dispenser image data are identified;

[0031] Step S202, three-dimensional positioning includes calibration processing and conversion processing;

[0032] Step S202.1, calibration processing. The internal parameters and distortion coefficients of the camera are obtained through camera calibration, and distortion correction is performed until the geometric shape of the dispenser in the image is consistent with the actual one;

[0033] Step S202.2, conversion processing. The PnP algorithm is used to convert the coordinates of the edge feature points on the dispenser image data plane into coordinates in the three-dimensional world coordinate system;

[0034] Step S202.3: Obtain the three-dimensional coordinate data of the injector image data based on Step S202.1 and Step S202.2;

[0035] Step S203: Pose analysis. Based on the three-dimensional coordinate positioning data, use the EPnP algorithm to solve the rotation matrix and translation vector of the injector relative to the camera, thereby determining the pose of the injector in three-dimensional space;

[0036] Step S203.1: Input processing. Through the three-dimensional point set to represent the position of the injector in the world coordinate system, where , and the corresponding two-dimensional image point set represents the projection of the three-dimensional point on the camera image plane;

[0037] Step S203.2: Solving process. First, solve the camera internal parameter matrix , specifically: In the formula, and are the focal lengths of the image in the and directions respectively, and are the coordinates of the image center;

[0038] The EPnP algorithm will construct an equation system based on the following projection relationship and solve and Specifically: In the formula, is the homogeneous coordinate of the image coordinate point, expressed as , is 's rotation matrix, representing the rotation of the camera relative to the world coordinate system, is 's translation vector, representing the displacement of the camera relative to the world coordinate system;

[0039] Step S203.3: Output processing. Output the rotation matrix and the translation vector , specifically: represents the rotation in three-dimensional space;

[0040] represents the translation of the camera relative to the world coordinate system;

[0041] Step S204: After obtaining the rotation matrix in Step S203, convert it to quaternion representation;

[0042] Step S204.1: Calculate the quaternion. Calculate each component of the quaternion, specifically: Wherein, is a rotation matrix, and its elements represent the element at the th row and the th column of the rotation matrix, represents the vector part in the quaternion, describing the direction of the rotation axis, is the scalar part in the quaternion, describing the cosine half-angle of the rotation angle,

[0043] Step S204.2, convert the output to obtain the quaternion , and use the obtained quaternion for subsequent attitude control of the robotic arm.

[0044] Combined with the first aspect, in some specific embodiments, plan the robotic arm gripping path based on the obtained attitude analysis result and the three-dimensional positioning result, specifically including:

[0045] Step S301, planning the robotic arm gripping path includes inputting the quaternion, the three-dimensional positioning result and the attitude analysis result, inverse kinematics solution, trajectory generation, and kinematic simulation;

[0046] Step S302, input the quaternion, the three-dimensional positioning result and the attitude analysis result;

[0047] Input the quaternion, and the quaternion is used to describe the rotation information of the applicator in three-dimensional space;

[0048] Input the three-dimensional positioning result , and the three-dimensional positioning result is used to describe the position coordinates of the applicator in the world coordinate system;

[0049] Input the attitude analysis result, and the attitude analysis result includes the direction of the rotation axis of the applicator and its angle relative to the camera;

[0050] Step S303, the inverse kinematics solution includes establishing the kinematic model of the robotic arm and solving the joint angles;

[0051] Step S303.1, establish the kinematic model of the robotic arm, and use the D-H parameters to describe each joint and link of the robotic arm, specifically:

[0052] Wherein, is the rotation angle, is the twist angle of the link, is the length of the link, is the offset;

[0053] Step S303.2: Solve for each joint angle, and use the Jacobian matrix to solve for the joint angles that enable the end effector of the robotic arm to reach the position of the drug applicator. Specifically, In the formula, represents the relationship between the position change of the end effector and the change in the joint angles of the robotic arm, are the angles of each joint;

[0054] Step S304: Trajectory generation includes attitude interpolation and smooth path planning;

[0055] Step S304.1: Attitude interpolation. Use the quaternion interpolation method to generate a smooth rotation path between the starting and target attitudes, and perform interpolation based on the quaternion values obtained in S204.2. Specifically, In the formula, is the angle between the quaternions, and are the starting and target quaternions, is the interpolation factor, The value range of ;

[0056] Step S304.2: Smooth path planning. Based on the obtained attitude interpolation results and the joint angles solved by inverse kinematics, generate a smooth grasping path through B-spline interpolation. Specifically, In the formula, represents the spatial position of the robotic arm at parameter time, represents the control points, and each control point corresponds to the three-dimensional coordinates of the robotic arm at a specific position, represents the th order B-spline basis function, which represents the influence degree of this control point on the curve shape at parameter ;

[0057] Among them, by using B-spline interpolation, the movement process of the robotic arm can be made stable and continuous, ensuring that the path meets the grasping accuracy while avoiding exceeding the joint angle limits of the robotic arm;

[0058] Step S305: Kinematic simulation. After the path is generated, perform dynamic simulation to check whether the path meets the speed and acceleration limits of the robotic arm. Use the Newton-Euler method for dynamic analysis to determine the joint torques of the robotic arm during movement. Specifically,

[0059] By performing recursive operations on each joint, the forces and torques of each joint of the robotic arm are solved. The robotic arm is regarded as a connection system of multiple rigid bodies, and the motion states of each joint are derived through the relationship between force and acceleration; In the formula, represents the resultant force of the th link, represents the mass of the th link, represents the acceleration of the th link, represents the torque of the th joint, represents the moment of inertia of the th link, represents the angular acceleration, represents the angular velocity, and represent the external force and external torque respectively;

[0060] Among them, by selecting the above method in step S305, it can ensure the smooth operation of the robotic arm during actual operation and avoid generating excessive impact force.

[0061] Step S306: Based on the angle changes and attitude rotation information of each joint of the robotic arm obtained in steps S301 to S305, generate a complete motion instruction.

[0062] Combined with the first aspect, in some specific embodiments, based on the planned grasping path, the robotic arm grasps the applicator and transports it to the mechanical turntable for ultrasonic welding treatment, and monitors the image of the applicator during the welding process, specifically including:

[0063] Step S401: Generate and issue a path instruction. Based on the grasping path generated in step S300 and the angle change information of each joint of the robotic arm, convert the data into specific motion instructions, and the motion instructions include joint angle instructions, speed and acceleration instructions, and attitude adjustment instructions;

[0064] Among them, the joint angle instruction is used to indicate the target angle change of each joint of the robotic arm during the grasping process;

[0065] The speed and acceleration instructions are used to set the speed and acceleration during the motion of the robotic arm to ensure smoothness during the path execution and avoid generating excessive impact or vibration;

[0066] The attitude adjustment instruction is based on the quaternion attitude information to adjust the attitude of the end effector of the robotic arm to ensure the correct direction and angle of the applicator during the grasping process;

[0067] Step S402. Manipulator motion control. When the manipulator executes a motion instruction, real-time feedback control is performed based on built-in sensors.

[0068] The sensors include a position sensor and a torque sensor. Based on real-time feedback, it is ensured that the manipulator stays on the planned path during motion. If a deviation between the joint angle and the preset instruction is detected during execution, error correction is automatically performed to adjust the joint angle to ensure that the manipulator accurately reaches the target position.

[0069] Step S403. Execute the grasping action. When the end of the manipulator reaches the position of the dispenser, the end effector executes the grasping action.

[0070] Among them, the opening and closing of the fixture are controlled by a servo motor. The specific steps are as follows:

[0071] A1. Grasping positioning. Based on the result of pose analysis, the pose of the end effector is adjusted to align it with the grasping point of the dispenser.

[0072] A2. Force control adjustment. During the closing process of the fixture, the force sensor is used to adjust the clamping force in real time to ensure an appropriate clamping force is applied to the dispenser to avoid damage.

[0073] Step S404. Transport to the mechanical turntable. After grasping is completed, the manipulator moves the dispenser to the mechanical turntable according to the planned path and releases the dispenser at a predetermined position, preparing for ultrasonic welding.

[0074] Step S405. Ultrasonic welding. When the manipulator accurately positions the dispenser on the convex engaging block of the welding table, the ultrasonic welding head applies pressure and transmits high-frequency vibration energy at the welding position of the dispenser, causing local melting of the welding surface of the dispenser to achieve welding.

[0075] Step S406. Image monitoring. An industrial camera is installed above the mechanical turntable for welding to monitor the state of the dispenser during the welding process.

[0076] Among them, the camera is connected to the system and can collect image data during the welding process in real time.

[0077] Step S407. Welding quality inspection. Through image monitoring and edge detection algorithm, the weld seam of the welded dispenser is detected to identify whether there are any defects in the weld seam.

[0078] Among them, the Canny edge detection algorithm is selected for the edge detection algorithm. After edge detection is completed, the extracted weld seam contour is compared with the ideal weld seam standard to detect whether there are any defects in the weld seam, and the continuity and uniformity of the weld seam are respectively detected.

[0079] Weld continuity is detected by analyzing whether the edges are continuous to check for weld fractures or porosity.

[0080] Weld uniformity is judged by measuring the change in weld width to determine whether the welding is uniform. In the formula, is the weld width, and are the coordinates of the upper and lower edge points of the weld at the position respectively, and is the number of edge points.

[0081] If there is no abnormality, the robotic arm confirms the welding quality through the image monitoring system. If the inspection is qualified, it sends a completion signal to the control system and transfers the dispenser to the next station.

[0082] If the inspection is unqualified, the abnormal information is recorded, the welding position is readjusted and an alarm signal is sent to prompt the operator to check and perform manual intervention.

[0083] Among them, in step S400, the robotic arm precisely grabs and transports the dispenser to the welding position by receiving a path planning instruction. During the welding process, the welding quality and position status are detected in real time through the image monitoring system to ensure the stability and reliability of the welding process. Finally, the robotic arm makes feedback adjustments based on the results of the image monitoring or transports the welded dispenser to the next process.

[0084] Combined with the first aspect, in some specific embodiments, path planning is performed on the robotic arm at the other end based on the image data of the welded dispenser, which specifically includes:

[0085] Step S501, Image acquisition and analysis: After welding is completed, the image data of the welded dispenser is obtained through an industrial camera, and feature extraction and analysis are performed on the image data, including identifying the welding state, position, and spatial attitude of the dispenser relative to the robotic arm at the other end.

[0086] Step S502, Based on the position and attitude of the welded dispenser, inverse kinematics is used to perform path planning on the robotic arm at the other end. The method of path planning is the same as the process in step S300, including inverse kinematics solution based on image data and B-spline interpolation to generate a smooth path, specifically referring to the specific process in steps S301 to S306.

[0087] Step S503, Through path planning, a robotic arm motion instruction is generated and output to the robotic arm at the other end, enabling the robotic arm to reach the position of the dispenser and complete subsequent gripping and transfer operations.

[0088] In combination with the first aspect, in some specific embodiments, the robotic arm at the other end picks up and transfers the processed drug dispenser to the conveyor belt based on the planned path, and the conveyor belt conveys the drug dispenser at a constant speed. Specifically, it includes:

[0089] Step S601: Picking operation. The picking operation in this step is the same as the picking operation method in step S400. For the specific description of the picking process, refer to step S400;

[0090] Step S602: Transfer processing. After the robotic arm at the other end completes the picking, it moves the drug dispenser to the designated position on the conveyor belt based on the planned path. The robotic arm transports the drug dispenser to the conveyor belt based on the posture analysis result and path interpolation data;

[0091] Step S603: Constant-speed conveying. After the robotic arm at the other end reaches the designated position above the conveyor belt, it controls the fixture to gradually loosen to ensure that the drug dispenser lands steadily on the conveyor belt. After the drug dispenser lands on the conveyor belt, the conveyor belt starts at a constant speed and conveys the drug dispenser to the next process position based on the preset speed.

[0092] In the second aspect, the present application further provides a robotic arm positioning picking and placing system for ultrasonic welding of drug dispensers, including the following modules:

[0093] A preprocessing module for acquiring and preprocessing the image data of the drug dispenser based on the conveyor belt conveying the drug dispenser;

[0094] An image processing module for extracting features based on the preprocessed image data of the drug dispenser, and performing three-dimensional positioning and posture analysis on the drug dispenser based on the extracted features;

[0095] A first path planning module for planning the picking path of the robotic arm based on the obtained posture analysis result and three-dimensional positioning result;

[0096] An ultrasonic welding module for the robotic arm to pick up the drug dispenser and transport it to the mechanical turntable based on the planned picking path for ultrasonic welding processing, and monitoring the image of the drug dispenser during the welding process;

[0097] A second path planning module for path planning of the robotic arm at the other end based on the image data of the drug dispenser after welding;

[0098] A transfer module for the robotic arm at the other end to pick up and transfer the processed drug dispenser to the conveyor belt based on the planned path, and the conveyor belt conveys the drug dispenser at a constant speed.

[0099] Compared with the prior art, the present invention has the following beneficial effects:

[0100] Through precise preprocessing of image data and pose analysis, the present invention realizes three-dimensional positioning of the applicator, optimizes the grasping and welding paths of the robotic arm through path planning to ensure the smoothness and accuracy of the welding process, and conducts real-time detection and feedback on the welding quality through an image monitoring system, effectively improving the welding quality and production efficiency, enhancing the automation level and operation accuracy of the overall method, and significantly reducing manual intervention and the defective product rate.

[0101] To make the above objects, features, and advantages of the present application more clearly understandable, the following specifically gives preferred embodiments and, in conjunction with the accompanying drawings, provides detailed descriptions as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required to be used in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other relevant drawings can also be obtained based on these drawings.

[0103] Figure 1 It is a flowchart of a method for positioning and picking up a robotic arm for ultrasonic welding of an applicator provided by an embodiment of the present application.

[0104] Figure 2 It is a schematic diagram of a system for positioning and picking up a robotic arm for ultrasonic welding of an applicator provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0105] The following will describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application.

[0106] Please refer to Figure 1 , Figure 1 It is a flowchart of a method for positioning and picking up a robotic arm for ultrasonic welding of an applicator provided by an example of the present application.

[0107] In this embodiment, a method for positioning and picking up a robotic arm for ultrasonic welding of an applicator may include step S100, step S200, step S300, step S400, step S500, and step S600.

[0108] Step S100: Based on the conveyor belt transporting the applicator, obtain the image data of the applicator and preprocess it.

[0109] Here, it specifically includes:

[0110] Step S101: Obtaining the image data of the applicator in the conveying state includes feeding processing and shooting processing;

[0111] Step S101.1: Feeding process. Manually place the pre-assembled drug dispenser at the conveyor belt, and the conveyor belt transports it at a constant speed.

[0112] Step S101.2: Obtain the image data of the drug dispenser in the conveying state. By presetting a shooting area above the conveying area of the drug dispenser and installing an industrial camera above the shooting area, take pictures of the drug dispenser to obtain continuous image data.

[0113] Step S102: Preprocessing of the drug dispenser image data. The preprocessing includes image denoising, grayscale processing, edge detection, and image enhancement.

[0114] Step S102.1: Image denoising. Use a Gaussian filter combined with median filtering to smooth the image and obtain the denoised image data of the drug dispenser.

[0115] Step S102.2: Grayscale processing. Convert the color image after denoising processing into a grayscale image, and use the weighted average method to perform grayscale processing on the weights of the three channels to obtain the grayscale processed image data of the drug dispenser.

[0116] Step S102.3: Edge detection. Use the Canny edge detection algorithm to extract the edges of the grayscale processed image and highlight the contour of the drug dispenser image data.

[0117] Step S102.4: Image enhancement. Use histogram equalization to enhance the contrast of the image and enhance the contour of the drug dispenser image data.

[0118] Step S102.5: Based on Steps S102.1 to S102.4, obtain the preprocessed image data of the drug dispenser.

[0119] Step S200: Based on the preprocessed image data of the drug dispenser, perform feature extraction, and perform three-dimensional positioning and attitude analysis on the drug dispenser based on the extracted features.

[0120] Here, it specifically includes:

[0121] Step S201: Feature extraction. Based on the detection results obtained in Step S102.3, extract the complete contour information in the drug dispenser image data and identify the positions of the edge feature points of the drug dispenser image data.

[0122] Step S202: Three-dimensional positioning includes calibration processing and conversion processing.

[0123] Step S202.1: Calibration processing. Obtain the internal parameters and distortion coefficients of the camera through camera calibration, perform distortion correction, and correct until the geometric shape of the drug dispenser in the image is consistent with the actual situation.

[0124] Step S202.2. Transformation processing: Use the PnP algorithm to convert the coordinates of the edge feature points on the data plane of the injector image into coordinates in the three-dimensional world coordinate system;

[0125] Step S202.3. Based on Steps S202.1 and S202.2, obtain the three-dimensional coordinate data of the injector image data;

[0126] Step S203. Pose analysis: Based on the three-dimensional coordinate positioning data, use the EPnP algorithm to solve the rotation matrix and translation vector of the injector relative to the camera, thereby determining the pose of the injector in three-dimensional space;

[0127] Step S203.1. Input processing: Through the three-dimensional point set indicating the position of the injector in the world coordinate system, where , and the corresponding two-dimensional image point set represents the projection of the three-dimensional point on the camera image plane;

[0128] Step S203.2. Solution processing: First, solve the camera internal parameter matrix , specifically: In the formula, and are the focal lengths of the image in the and directions respectively, and and are the coordinates of the image center;

[0129] The EPnP algorithm will construct an equation set based on the following projection relationship and solve and Specifically: In the formula, is the homogeneous coordinate of the image coordinate point, expressed as , is 's rotation matrix, representing the rotation of the camera relative to the world coordinate system, is 's translation vector, representing the displacement of the camera relative to the world coordinate system;

[0130] Step S203.3. Output processing: Output the rotation matrix and the translation vector , specifically: represents the rotation in three-dimensional space;

[0131] represents the translation of the camera relative to the world coordinate system;

[0132] Step S204: After obtaining the rotation matrix in step S203, convert it into quaternion representation;

[0133] Step S204.1: Calculate the quaternion. By calculating each component of the quaternion, specifically: In the formula, is the rotation matrix, and its element represents the -th row and -th column element of the rotation matrix. represents the vector part in the quaternion, describing the direction of the rotation axis. is the scalar part in the quaternion, describing the cosine half-angle of the rotation angle.

[0134] Step S204.2: Convert the output to obtain the quaternion , and use the obtained quaternion for subsequent attitude control of the robotic arm.

[0135] Step S300: Plan the grasping path of the robotic arm based on the obtained attitude analysis result and 3D positioning result.

[0136] Here, it specifically includes:

[0137] Step S301: Planning the grasping path of the robotic arm includes inputting the quaternion, 3D positioning result and attitude analysis result, inverse kinematics solution, trajectory generation, and kinematic simulation;

[0138] Step S302: Input the quaternion, 3D positioning result and attitude analysis result;

[0139] Input the quaternion. The quaternion is used to describe the rotation information of the dispenser in 3D space;

[0140] Input the 3D positioning result , and the 3D positioning result is used to describe the position coordinates of the dispenser in the world coordinate system;

[0141] Input the attitude analysis result. The attitude analysis result includes the direction of the rotation axis of the dispenser and its angle relative to the camera;

[0142] Step S303: The inverse kinematics solution includes establishing the kinematic model of the robotic arm and solving the joint angles;

[0143] Step S303.1: Establish the kinematic model of the robotic arm. Use the D-H parameters to describe each joint and link of the robotic arm, specifically:

[0144] wherein is the rotation angle, is the torsional angle of the connecting rod, is the length of the connecting rod, is the offset;

[0145] Step S303.2, solve each joint angle, and use the Jacobian matrix to solve the joint angles that make the end of the robotic arm reach the position of the applicator Specifically: wherein represents the relationship between the position change of the end effector and the change of the robotic arm joint angle, are the angles of each joint;

[0146] Step S304, trajectory generation includes attitude interpolation and smooth path planning;

[0147] Step S304.1, attitude interpolation, use the quaternion interpolation method to generate a smooth rotation path between the starting and target attitudes, and perform interpolation based on the quaternion values obtained in S204.2. Specifically: wherein is the angle between the quaternions, and are the starting and target quaternions, is the interpolation factor, the value range of ;

[0148] Step S304.2, smooth path planning, based on the obtained attitude interpolation results and the joint angles solved by inverse kinematics, generate a smooth grasping path through B-spline interpolation. Specifically: wherein represents the spatial position of the robotic arm at parameter time, represents the control point, and each control point corresponds to the three-dimensional coordinates of the robotic arm at a specific position, represents the th control point's order B-spline basis function, which represents the influence degree of this control point on the curve shape at parameter ;

[0149] Step S305, kinematic simulation. After the path is generated, perform dynamic simulation to check whether the path meets the speed and acceleration limits of the robotic arm, and use the Newton-Euler method for dynamic analysis to determine the joint torques of the robotic arm during movement. Specifically:

[0150] By performing recursive operations on each joint, the forces and torques of each joint of the robotic arm are solved. The robotic arm is regarded as a connection system of multiple rigid bodies, and the motion states of each joint are deduced through the relationship between force and acceleration. In the formula, represents the resultant force of the th link, represents the mass of the th link, represents the acceleration of the th link, represents the torque of the th joint, represents the moment of inertia of the th link, represents the angular acceleration, represents the angular velocity, and represent the external force and external torque respectively;

[0151] Step S306: Based on the angle changes and attitude rotation information of each joint of the robotic arm obtained in steps S301 to S305, generate a complete motion instruction.

[0152] Step S400: Based on the planned gripping path, the robotic arm grips the applicator and transports it to the mechanical turntable for ultrasonic welding treatment, and monitors the image of the applicator during the welding process.

[0153] Here, it specifically includes:

[0154] Step S401: Generate and issue a path instruction. Based on the gripping path generated in step S300 and the angle change information of each joint of the robotic arm, convert the data into specific motion instructions, which include joint angle instructions, speed and acceleration instructions, and attitude adjustment instructions;

[0155] Among them, the joint angle instruction is used to indicate the target angle change of each joint of the robotic arm during the gripping process;

[0156] The speed and acceleration instructions are used to set the speed and acceleration during the motion of the robotic arm to ensure smoothness during the path execution and avoid excessive impact or vibration;

[0157] The attitude adjustment instruction is based on the quaternion attitude information to adjust the attitude of the end effector of the robotic arm to ensure the correct direction and angle of the applicator during the gripping process;

[0158] Step S402: Motion control of the robotic arm. When the robotic arm executes the motion instruction, it performs real-time feedback control based on the built-in sensors;

[0159] The sensor includes a position sensor and a torque sensor, which ensures that the robotic arm stays on the planned path during movement based on real-time feedback. If a deviation between the joint angle and the preset command is detected during execution, error correction is automatically performed to adjust the joint angle to ensure that the robotic arm accurately reaches the target position;

[0160] Step S403: Perform the clamping action. When the end of the robotic arm reaches the position of the applicator, the end clamp performs the clamping action;

[0161] Step S404: Transport to the mechanical turntable. After the clamping is completed, the robotic arm moves the applicator to the mechanical turntable according to the planned path and releases the applicator at a predetermined position to prepare for ultrasonic welding;

[0162] Step S405: Ultrasonic welding. When the robotic arm accurately positions the applicator on the convex engaging block of the welding table, the ultrasonic welding head applies pressure and transmits high-frequency vibration energy at the welding position of the applicator, causing local melting of the welding surface of the applicator to achieve welding;

[0163] Step S406: Image monitoring. An industrial camera is installed above the mechanical turntable for welding to monitor the state of the applicator during the welding process;

[0164] Step S407: Welding quality inspection. Through image monitoring and edge detection algorithms, the weld seam of the welded applicator is inspected to identify whether there are any defects in the weld seam;

[0165] Among them, the Canny edge detection algorithm is selected for the edge detection algorithm. After the edge detection is completed, the extracted weld seam contour is compared with the ideal weld seam standard to detect whether there are any defects in the weld seam, and the continuity and uniformity of the weld seam are respectively detected;

[0166] Weld seam continuity. By analyzing whether the edge is continuous, it is detected whether there are weld seam breaks or pores;

[0167] Weld seam uniformity. By measuring the change in the weld seam width, it is judged whether the welding is uniform; In the formula, is the weld seam width, and are the coordinates of the upper and lower edge points of the weld seam at the position respectively, is the number of edge points.

[0168] If there is no abnormality, the robotic arm confirms the welding quality through the image monitoring system. If the inspection is qualified, a completion signal is sent to the control system, and the applicator is transferred to the next station;

[0169] If the detection fails, record the abnormal information, readjust the welding position, and send an alarm signal to prompt the operator to check and perform manual intervention through the alarm signal.

[0170] Step S500: Based on the image data of the medicator after welding, perform path planning for the robotic arm at the other end.

[0171] Specifically, it includes:

[0172] Step S501: Image acquisition and analysis. After welding is completed, use an industrial camera to obtain the image data of the medicator after welding, and perform feature extraction and analysis on the image data, including identifying the welding state, position, and spatial attitude of the medicator relative to the robotic arm at the other end.

[0173] Step S502: Based on the position and attitude of the medicator after welding, use inverse kinematics to perform path planning for the robotic arm at the other end. The path planning method is the same as the process in Step S300, including inverse kinematics solution based on image data and B-spline interpolation to generate a smooth path. Specifically refer to the specific process in Steps S301 to S306.

[0174] Step S503: Through path planning, generate a robotic arm motion instruction and output it to the robotic arm at the other end, so that the robotic arm reaches the position of the medicator and completes the subsequent clamping and transfer operations.

[0175] Step S600: The robotic arm at the other end clamps and transfers the processed medicator to the conveyor belt based on the planned path, and the conveyor belt conveys the medicator at a constant speed.

[0176] Specifically, it includes:

[0177] Step S601: Clamping operation. The clamping operation in this step is the same as the clamping operation method in Step S400. Specifically refer to the description of the clamping process in Step S400.

[0178] Step S602: Transfer processing. After the robotic arm at the other end completes clamping, move the medicator to the designated position on the conveyor belt based on the planned path. The robotic arm transports the medicator to the conveyor belt based on the attitude analysis result and path interpolation data.

[0179] Step S603: Constant-speed conveying. After the robotic arm at the other end reaches the designated position above the conveyor belt, control the fixture to gradually loosen to ensure that the medicator lands steadily on the conveyor belt. After the medicator lands on the conveyor belt, the conveyor belt starts at a constant speed and conveys the medicator to the next process position based on the preset speed.

[0180] In actual operation, first, the applicator is conveyed by a conveyor belt. The pre-assembled applicator is manually placed on the conveyor belt, which transports it to the shooting area at a constant speed. An industrial camera is installed above the conveyor belt to capture consecutive images of the applicator. The shooting frequency of the camera is adjusted according to the speed of the conveyor belt and the size of the applicator to ensure clear and complete images.

[0181] Subsequently, the acquired image data is processed. The preprocessing includes: removing noise through Gaussian filtering and median filtering, grayscale processing of the image, edge detection using the Canny algorithm to extract the contour of the applicator, and enhancing the image contrast through histogram equalization to generate preprocessed image data.

[0182] After preprocessing, based on the results of edge detection, the complete contour of the applicator is extracted, and its edge feature points are identified. The PnP algorithm is used to convert the extracted two-dimensional feature points into three-dimensional coordinates. Through camera calibration and correction, the position of the applicator in space is obtained. Then, the EPnP algorithm is used to calculate the rotation matrix and translation vector of the applicator to describe its pose in three-dimensional space. The rotation matrix is converted into a quaternion through a formula to describe the rotation direction and angle of the applicator.

[0183] After that, based on the obtained three-dimensional positioning results and pose information, the quaternion and position information are input. Then, the kinematic model of the robotic arm is established according to the D-H parameters, and the joint angles of the robotic arm are solved through the Jacobian matrix to ensure that the end effector of the robotic arm can accurately reach the position of the applicator. The B-spline interpolation method is used to generate a continuous and smooth grasping path based on the solved joint angles and smooth path planning. Then, the Newton-Euler method is used for dynamic simulation to check whether the path meets the speed and acceleration limits of the robotic arm to ensure stable operation of the robotic arm.

[0184] At this time, the robotic arm will grasp the applicator based on the planned path and accurately transfer it to the mechanical turntable. During the process, the joint angles and poses are adjusted in real time. On the mechanical turntable, the robotic arm accurately positions the applicator, the ultrasonic welding head applies pressure and transmits high-frequency vibration energy to weld the contact surface of the applicator. An industrial camera installed on the welding table monitors the welding process in real time to detect the weld quality, and an image processing algorithm is used to judge whether the welding is qualified.

[0185] After welding is completed, the industrial camera acquires the image data of the welded applicator and analyzes it, including the welding state, position, and pose. Based on the analysis results, the robotic arm at the other end solves the path through inverse kinematics and uses B-spline interpolation to generate a smooth path to ensure the stability of the grasping process.

[0186] The robotic arm at the other end precisely picks up the processed drug dispenser based on the planned path and transfers it to the conveyor belt. After the robotic arm releases the drug dispenser, the conveyor belt transports the drug dispenser to the next process at a constant speed.

[0187] Please refer to Figure 2 , Figure 2 which is a schematic diagram of a robotic arm positioning and picking system for ultrasonic welding of a drug dispenser in this application.

[0188] A robotic arm positioning and picking system for ultrasonic welding of a drug dispenser includes the following modules:

[0189] A pretreatment module for obtaining and preprocessing the image data of the drug dispenser based on the conveyor belt transporting the drug dispenser.

[0190] An image processing module for extracting features based on the preprocessed image data of the drug dispenser, and performing three-dimensional positioning and pose analysis on the drug dispenser based on the extracted features.

[0191] A first path planning module for planning the robotic arm picking path based on the obtained pose analysis result and three-dimensional positioning result.

[0192] An ultrasonic welding module for the robotic arm to pick up the drug dispenser based on the planned picking path and transport it to the mechanical turntable for ultrasonic welding processing, and monitoring the image of the drug dispenser during the welding process.

[0193] A second path planning module for path planning of the robotic arm at the other end based on the image data of the drug dispenser after welding.

[0194] A transfer module for the robotic arm at the other end to pick up and transfer the processed drug dispenser to the conveyor belt based on the planned path, and the conveyor belt transports the drug dispenser at a constant speed.

[0195] As mentioned above, it is only the specific implementation manner of the present invention, but the protection scope of the embodiments of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the embodiments of the present invention should not easily think of changes or substitutions, and all should be covered within the protection scope of the embodiments of the present invention.

Claims

1. A robotic arm positioning and picking method for ultrasonic welding of a drug dispenser, characterized in that, The following are included: S100. Based on the conveyor belt to transport the applicator, obtain the image data of the applicator and preprocess it; S200. Extract features based on the preprocessed applicator image data, and perform three-dimensional positioning and pose analysis on the applicator based on the extracted features; S300. Plan the clamping path of the robotic arm based on the obtained pose analysis result and three-dimensional positioning result; S400. Based on the planned clamping path, the robotic arm clamps the applicator and transports it to the mechanical turntable for ultrasonic welding treatment, and monitors the image of the applicator during the welding process. Specifically, it includes: S401. Generate and issue path instructions. Based on the clamping path generated in S300 and the change information of each joint angle of the robotic arm, convert the data into specific motion instructions. The motion instructions include joint angle instructions, speed and acceleration instructions, and pose adjustment instructions; S402. Robotic arm motion control. When the robotic arm executes the motion instructions, it performs real-time feedback control based on the built-in sensors; S403. Execute the clamping action. When the end of the robotic arm reaches the position of the applicator, execute the clamping action through the end fixture; S404. Transport to the mechanical turntable. After the clamping is completed, the robotic arm moves the applicator to the mechanical turntable according to the planned path and releases the applicator at a predetermined position to prepare for ultrasonic welding; S405. Ultrasonic welding. When the robotic arm accurately positions the applicator on the convex engaging block of the welding table, the ultrasonic welding head applies pressure and transmits high-frequency vibration energy at the welding position of the applicator, causing the local melting of the welding surface of the applicator to achieve welding; S406. Image monitoring. Install an industrial camera above the mechanical turntable for welding to monitor the state of the applicator during the welding process; S407. Welding quality inspection. Through image monitoring and edge detection algorithms, detect the weld seam of the welded applicator to identify whether there are any defects in the weld seam; If there is no abnormality, the robotic arm confirms the welding quality through the image monitoring system. If the inspection is qualified, it sends a completion signal to the control system and transfers the applicator to the next workstation; If the inspection is unqualified, record the abnormal information, readjust the welding position and send an alarm signal, and prompt the operator to check and perform manual intervention through the alarm signal; S500. Based on the image data of the welded applicator, plan the path for the robotic arm at the other end; S600. The robotic arm at the other end clamps and transfers the processed applicator to the conveyor belt based on the planned path, and the conveyor belt transports the applicator at a constant speed.

2. The robotic arm positioning and picking method for ultrasonic welding of a drug dispenser according to claim 1, characterized in that, Based on the conveyor belt to transport the applicator, obtain the image data of the applicator and preprocess it. Specifically, it includes: S101. Obtain the image data of the applicator in the conveying state, including feeding treatment and shooting treatment; S101.

1. Feeding treatment. Manually place the pre-assembled applicator on the conveyor belt, and the conveyor belt transports it at a constant speed; S101.

2. Obtain the image data of the applicator in the conveying state. By presetting a shooting area above the applicator conveying area and installing an industrial camera above the shooting area, shoot the applicator to obtain continuous image data; S102. Preprocess the image data of the dispenser. The preprocessing includes image denoising, grayscale processing, edge detection, and image enhancement. S102.

1. Image denoising: Smooth the image using a Gaussian filter combined with median filtering to obtain the denoised image data of the dispenser. S102.

2. Grayscale processing: Convert the color image after denoising into a grayscale image. Use the weighted average method to perform grayscale processing with the weights of the three channels to obtain the grayscale processed image data of the dispenser. S102.

3. Edge detection: Use the Canny edge detection algorithm to extract the edges of the image after grayscale processing, highlighting the contour of the dispenser image data. S102.

4. Image enhancement: Use histogram equalization to enhance the contrast of the image and enhance the contour of the dispenser image data. S102.

5. Based on S102.1 to S102.4, obtain the preprocessed image data of the dispenser.

3. A robotic arm positioning and picking method for ultrasonic welding of a drug dispenser, characterized in that Perform feature extraction based on the preprocessed image data of the dispenser, and perform three-dimensional positioning and pose analysis on the dispenser based on the extracted features. Specifically, it includes: S201. Feature extraction: Based on the detection results obtained in S102.3, extract the complete contour information in the dispenser image data and identify the positions of the edge feature points in the dispenser image data. S202. Three-dimensional positioning includes calibration processing and conversion processing. S202.

1. Calibration processing: Obtain the internal parameters and distortion coefficients of the camera through camera calibration, and perform distortion correction until the geometric shape of the dispenser in the image is consistent with the actual one. S202.

2. Conversion processing: Use the PnP algorithm to convert the coordinates of the edge feature points on the dispenser image data plane into coordinates in the three-dimensional world coordinate system. S202.

3. Based on S202.1 and S202.2, obtain the three-dimensional coordinate data of the dispenser image data. S203. Pose analysis: Based on the three-dimensional coordinate positioning data, use the EPnP algorithm to solve the rotation matrix and translation vector of the dispenser relative to the camera, thereby determining the pose of the dispenser in three-dimensional space. S203.

1. Input processing, through a three-dimensional point set representing the position of the applicator in the world coordinate system, where , and the corresponding two-dimensional image point set represents the projection of the three-dimensional point on the camera image plane; S203.

2. Solving process: First, solve the camera internal parameter matrix , specifically as follows: In the formula, and are the focal lengths of the image in the and directions respectively, and and are the coordinates of the image center; The EPnP algorithm constructs an equation system based on the following projection relationships and solves it by minimizing the reprojection error and Specifically: In the formula, is the homogeneous coordinate of the image coordinate point, expressed as , is rotation matrix, representing the rotation of the camera relative to the world coordinate system, is translation vector, representing the displacement of the camera relative to the world coordinate system; S203.

3. Output processing, output the rotation matrix and the translation vector , specifically: Represents a rotation in three-dimensional space; Indicates the translation of the camera relative to the world coordinate system; S204. After obtaining the rotation matrix in S203, convert it into quaternion representation. S204.

1. Calculate the quaternion: Calculate each component of the quaternion. Specifically: In the formula, is the rotation matrix, and its elements represent the element in the th row and th column of the rotation matrix, represents the vector part in the quaternion and describes the direction of the rotation axis, is the scalar part in the quaternion and describes the cosine half-angle of the rotation angle, S204.

2. Convert the output to obtain a quaternion and use the obtained quaternion for subsequent attitude control of the robotic arm.

4. A robotic arm positioning and picking method for ultrasonic welding of a drug dispenser, characterized in that Plan the grasping path of the robotic arm based on the obtained pose analysis results and three-dimensional positioning results. Specifically, it includes: S301. Planning the grasping path of the robotic arm includes inputting the quaternion, three-dimensional positioning results, and pose analysis results, inverse kinematics solution, trajectory generation, and kinematic simulation. S302. Input the quaternion, three-dimensional positioning results, and pose analysis results. Input quaternion, quaternion Used to describe the rotation information of the applicator in three-dimensional space; Input three-dimensional positioning result , and the three-dimensional positioning result is used to describe the position coordinates of the applicator in the world coordinate system; Input the pose analysis results, which include the direction of the rotation axis of the dispenser and its angle relative to the camera. S303. Inverse kinematics solution includes establishing the kinematic model of the robotic arm and solving the joint angles. S303.

1. Establish the kinematic model of the robotic arm: Use the D-H parameters to describe each joint and link of the robotic arm. Specifically: In the formula, is the rotation angle, is the torsion angle of the connecting rod, is the length of the connecting rod, is the offset; S303.

2. Solve the angles of each joint, and use the Jacobian matrix to solve the angles of each joint so that the end of the robotic arm reaches the position of the applicator. Specifically, they are as follows: In the formula, represents the relationship between the position change of the end effector and the joint angle change of the robotic arm, are the angles of each joint; S304. Trajectory generation includes pose interpolation and smooth path planning. S304.

1. Pose interpolation. Using the quaternion interpolation method, a smooth rotation path is generated between the starting and target poses. Interpolation is performed based on the quaternion values obtained in S204.

2. Specifically: Wherein, is the angle between quaternions, and are the starting and target quaternions, is the interpolation factor, The value range of is 1); S304.

2. Smooth path planning. Based on the obtained pose interpolation results and the joint angles solved by inverse kinematics, a smooth grasping path is generated through B-spline interpolation. Specifically: In the formula, represents the spatial position of the robotic arm at parameter moment, represents the control point, and each control point corresponds to the three-dimensional coordinates of the robotic arm at a specific position, represents the th order B-spline basis function, which represents the degree of influence of this control point on the curve shape at parameter ; S305. Kinematic simulation. After the path is generated, dynamic simulation is performed to check whether the path meets the speed and acceleration limits of the robotic arm. The Newton-Euler method is used for dynamic analysis to determine the joint torques of the robotic arm during movement. Specifically: By performing recursive operations on each joint, the forces and torques of each joint of the robotic arm are solved. The robotic arm is regarded as a connection system of multiple rigid bodies, and the motion states of each joint are deduced through the relationship between force and acceleration. In the formula, represents the resultant force of the th connecting rod, represents the mass of the th connecting rod, represents the acceleration of the th connecting rod, represents the torque of the th joint, represents the moment of inertia of the th connecting rod, represents the angular acceleration, represents the angular velocity, and respectively represent the external force and the external torque; S306. Based on S301 to S305, the angle changes and pose rotation information of each joint of the robotic arm are obtained, and a complete motion instruction is generated.

5. A robotic arm positioning and picking method for ultrasonic welding of a drug delivery device according to claim 4, characterized in that, Based on the image data of the medicator after welding, path planning is performed on the robotic arm at the other end. Specifically, it includes: S501. Image acquisition and analysis. After welding is completed, the image data of the medicator after welding is obtained through an industrial camera, and feature extraction and analysis are performed on the image data, including identifying the welding state, position, and spatial pose of the medicator relative to the robotic arm at the other end. S502. Based on the position and pose of the medicator after welding, inverse kinematics is used to perform path planning on the robotic arm at the other end. The method of path planning is the same as the process in S300, including inverse kinematics solution based on image data and B-spline interpolation to generate a smooth path. Specifically refer to the specific process in S301 to S306. S503. Through path planning, a robotic arm motion instruction is generated and output to the robotic arm at the other end, so that the robotic arm reaches the position of the medicator and completes the subsequent grasping and transfer operations.

6. A robotic arm positioning and picking method for ultrasonic welding of a drug delivery device according to claim 1, characterized in that, The robotic arm at the other end grasps and transfers the processed medicator to the conveyor belt based on the planned path. The conveyor belt conveys the medicator at a constant speed. Specifically, it includes: S601. Grasping operation. The grasping operation in this step is the same as the grasping operation method in S400. Specifically refer to the description of the grasping process in S400. S602. Transfer processing. After grasping is completed, the robotic arm at the other end moves the medicator to the designated position on the conveyor belt based on the planned path. The robotic arm transports the medicator to the conveyor belt based on the pose analysis results and path interpolation data. S603. Constant-speed conveying. After reaching the designated position above the conveyor belt, the robotic arm at the other end controls the fixture to gradually loosen to ensure that the medicator lands smoothly on the conveyor belt. After the medicator lands on the conveyor belt, the conveyor belt starts at a constant speed and conveys the medicator to the next process position based on the preset speed.

7. A robotic arm positioning and picking system for ultrasonic welding of a medicator, adopting a robotic arm positioning and picking method for ultrasonic welding of a medicator according to any one of claims 1-6, characterized in that, It includes the following modules: Preprocessing module, used to obtain the image data of the medicator and preprocess it based on the conveyor belt conveying the medicator. Image processing module, used to perform feature extraction based on the preprocessed image data of the medicator, and perform three-dimensional positioning and pose analysis on the medicator based on the extracted features. The first path planning module is used to plan the robotic arm gripping path based on the obtained attitude analysis result and the three-dimensional positioning result; The ultrasonic welding module is used to, based on the planned gripping path, the robotic arm grips the applicator and conveys it to the mechanical turntable for ultrasonic welding treatment, and performs image monitoring on the applicator during the welding process; The second path planning module is used to plan the path of the robotic arm at the other end based on the image data of the welded applicator; The transfer module is used for the robotic arm at the other end to grip and transfer the processed applicator to the conveyor belt based on the planned path, and the conveyor belt conveys the applicator at a constant speed.

Images