Metal handicraft visual point painting dispensing equipment control system and method

Abstract

This application relates to the field of automated motion control technology, and discloses a control system and method for a visual paint dispensing and gluing equipment for metal crafts. The system acquires standard target trajectory coordinates and workpiece surface morphology feature data; calculates inverse vector space bias data based on a preset interpolation direction gradient vector field, and superimposes it with the original coordinates to generate reconstructed target trajectory coordinates to counteract the sloping fluid slippage of the paint; maps the reconstructed target trajectory coordinates to a surface morphology feature matrix, extracts local relative undulation data and local concave-convex curvature feature data, and calculates and generates Z-axis follow-up height command parameters and dispensing volume pulse width command parameters, respectively; finally, it controls the actuator and dispensing valve to perform three-dimensional collaborative operation based on the above parameters. This invention achieves adaptive dynamic compensation of dispensing gap and dispensing volume, effectively solving the problems of sloping dispensing trajectory offset, paint stringing and dripping, and uneven coating thickness caused by microscopic undulations, thus improving the accuracy and stability of collaborative processing.

Description

Technical Field

[0001] This invention relates to the field of automated motion control technology, specifically to a control system and method for a visual painting and dispensing equipment for metal crafts. Background Technology

[0002] In the manufacturing process of metal crafts, surface painting or adhesive application is often required to enhance aesthetics or achieve specific surface treatments. Existing automated vision-based adhesive application equipment mostly uses standard two-dimensional discrete coordinates for path planning and position interpolation. However, most metal craft surfaces are not perfectly flat; they often exhibit slopes, microscopic undulations, and complex textures. Existing two-dimensional planar control methods have significant technical limitations when dealing with such complex workpiece shapes.

[0003] When applying paint to a sloping area of ​​a workpiece, the paint droplets, influenced by their own gravity and fluid properties, undergo fluid slippage before curing. Because existing control systems lack a pre-compensation mechanism for the physical laws governing fluid slippage on sloping surfaces, the final curing position of the paint deviates from the predetermined target trajectory. Simultaneously, due to the microscopic undulations on the workpiece surface, in an operation mode where the equipment maintains a fixed Z-axis height, the physical gap between the dispensing valve outlet and the workpiece surface fluctuates with these undulations. An excessively large gap can easily cause paint stringing, while an excessively small gap can lead to interference or collision between the outlet and the workpiece, or even squeezing and dripping. Furthermore, for areas with microscopic surface irregularities, existing equipment typically outputs the same volume of paint according to a set fixed pulse width, failing to adaptively adjust the fluid output volume based on local curvature characteristics. This results in insufficient filling of depressions and overflowing paint accumulation on convex areas, ultimately leading to uneven paint coating thickness on the surface of the craft, making it difficult to meet the high-precision appearance quality requirements of metal crafts. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a visual paint dispensing and coating equipment control system and method for metal crafts. This solves the problems of existing equipment causing paint to easily slip on inclined surfaces when processing metal crafts with undulating or textured surfaces, leading to the final curing position deviating from the target trajectory; fluctuations in the gap between the dispensing valve outlet and the workpiece surface easily causing paint stringing or dripping; and inconsistent paint layer thickness after curing due to a fixed dispensing amount in the microscopic uneven areas of the workpiece surface.

[0005] To address the above problems, the present invention provides the following technical solution: The first aspect of this invention provides a control system for a visual painting and adhesive application device for metal crafts, comprising: The data extraction module is used to acquire standard two-dimensional discrete target trajectory coordinate data and surface morphology feature data of the metal crafts to be processed; The trajectory reconstruction calculation module is used to calculate the inverse vector space bias data corresponding to the standard two-dimensional discrete target trajectory coordinate data based on the preset interpolation direction gradient vector field, and to superimpose the inverse vector space bias data with the standard two-dimensional discrete target trajectory coordinate data to generate the reconstructed target trajectory coordinates; The three-dimensional parametric coupling module is used to map the reconstructed target trajectory coordinates to the surface topography feature matrix, extract local relative undulation data and local concave-convex curvature feature data, and calculate and generate Z-axis follow-up height command parameters and dispensing volume pulse width command parameters respectively. The underlying collaborative execution module is used to control the actuator and dispensing valve to perform three-dimensional collaborative dispensing operations based on the reconstructed target trajectory coordinates, the Z-axis follow-up height command parameter, and the dispensing volume pulse width command parameter.

[0006] Furthermore, when calculating and generating the reconstructed target trajectory coordinates, the trajectory reconstruction calculation module specifically performs the following steps: retrieves the hand-eye calibration matrix and maps the standard two-dimensional discrete target trajectory coordinate data to the pixel reference system where the interpolation direction gradient vector field is located; obtains the corresponding local direction gradient vector in the pixel reference system; converts the local direction gradient vector into a physical direction gradient vector in the mechanical motion coordinate system according to the scaling and rotation parameters of the hand-eye calibration matrix; multiplies the physical direction gradient vector with a preset fluid slip compensation coefficient to generate the inverse vector space offset data defined in the mechanical motion coordinate system; and performs translation and superposition calculations on the inverse vector space offset data and the standard two-dimensional discrete target trajectory coordinate data to generate the reconstructed target trajectory coordinates.

[0007] Furthermore, when the three-dimensional parametric coupling module extracts local relative undulation data and calculates and generates Z-axis follow-up height command parameters, it is specifically used to: map the reconstructed target trajectory coordinates to the pixel reference system where the surface topography feature matrix is ​​located, and use the mapped pixel coordinates as an index to find the corresponding local relative undulation data in the surface topography feature matrix; multiply the local relative undulation data by a preset height mapping coefficient to convert it into a physical height offset; and superimpose or subtract the physical height offset from the global absolute reference height parameter, the Z-axis relative installation offset, and the preset dispensing working gap to generate the Z-axis follow-up height command parameter for the current interpolation cycle.

[0008] In a preferred embodiment of the present invention, when the three-dimensional parametric coupling module searches for the corresponding local relative undulation data in the surface topography feature matrix, it is specifically used to: employ a bilinear interpolation algorithm to extract the integer pixel values ​​around the mapped pixel coordinates and perform a weighted summation to calculate the corresponding local relative undulation data; if the mapped pixel coordinates are close to the effective boundary of the surface topography feature matrix, resulting in insufficient available adjacent pixels, the nearest neighbor interpolation algorithm is switched to extract the data.

[0009] Furthermore, the height mapping coefficient is obtained through a visual offline calibration program for the stepped blocks. The acquisition process is as follows: acquiring images of stepped blocks with known height differences, calculating the surface morphology feature matrix values ​​corresponding to each stepped region; using the known height difference as the fitting target, obtaining the height mapping coefficient used to scale the relative undulation values ​​to mechanical displacement through linear fitting or least squares fitting.

[0010] Furthermore, when the three-dimensional parametric coupling module extracts local concave-convex curvature feature data and calculates and generates the dispensing volume pulse width command parameter, it specifically performs the following: It uses a discrete difference template to calculate the two-dimensional Laplacian operator data corresponding to the coordinates of the reconstructed target trajectory in the surface topography feature matrix, and uses the two-dimensional Laplacian operator data as the local concave-convex curvature feature data characterizing the local concave-convex polarity and steepness of the workpiece surface; it multiplies the local concave-convex curvature feature data with a preset volume compensation gain to obtain the corresponding effective level duration increment of the dispensing valve; it adds the reference dispensing pulse width constant required for the dispensing valve to work with the effective level duration increment to generate the dispensing volume pulse width command parameter for the current interpolation cycle.

[0011] Furthermore, before calculating and generating the Z-axis follow-up height command parameter and the dispensing volume pulse width command parameter respectively, the three-dimensional parameter coupling module is also specifically used to: determine the validity of the gray-scale detection features at the mapped target coordinates and the two-dimensional Laplacian operator data; when the gray-scale value at the target coordinates is in the saturation range, the gray-scale gradient difference value exceeds the preset abnormal threshold, the magnitude of the local directional gradient vector exceeds the preset gradient threshold, or the two-dimensional Laplacian operator data exceeds the preset divergence threshold, the position is determined as an abnormal reflection point or an invalid detection point, and the neighborhood median filtering result, the neighborhood mean, or the data of the previous effective interpolation period are preferentially used for replacement.

[0012] Furthermore, when the underlying collaborative execution module controls the actuator and dispensing valve to perform three-dimensional collaborative dispensing operations, it specifically performs the following: It divides the reconstructed target trajectory coordinates with a set mechanical pulse equivalent constant, and rounds them to convert them into X-axis and Y-axis target pulse counts recognizable by the X-axis and Y-axis motors; it encapsulates the X-axis target pulse count, the Y-axis target pulse count, the Z-axis follow-up height command parameter, and the dispensing volume pulse width command parameter into a multi-axis synchronous data packet according to a set motion control protocol; it sends the multi-axis synchronous data packet to the underlying multi-axis servo driver and independent dispensing controller via an industrial fieldbus to control the actuator to perform planar position interpolation and vertical micro-motion tracking, and synchronously triggers the dispensing valve to output the corresponding effective high-level duration or duty cycle electrical signal during a set trigger window period.

[0013] In another preferred embodiment of the present invention, before the underlying collaborative execution module sends the multi-axis synchronous data packet to the multi-axis servo driver and the dispensing controller, it is further specifically used to: perform safety constraint judgment on the control parameters corresponding to the X-axis motor, Y-axis motor, Z-axis motor and dispensing valve; determine whether the reconstructed target trajectory coordinates are within the mechanical soft limit range of the X-axis and Y-axis, determine whether the Z-axis follow-up height command parameter is within the safe soft limit range of the Z-axis, and determine whether the dispensing volume pulse width command parameter is not lower than the minimum mechanical response time of the dispensing valve and not greater than the maximum allowable cycle time of the current system interpolation cycle; when any control parameter exceeds the corresponding safe range, perform truncation, alarm or suspension of execution processing.

[0014] A second aspect of the present invention provides a control method for a visual painting and dispensing equipment for metal crafts, applied to the aforementioned control system for the visual painting and dispensing equipment for metal crafts, comprising the following steps: Acquire standard two-dimensional discrete target trajectory coordinate data and surface morphology feature data of the metal craft to be processed; The inverse vector space bias data corresponding to the standard two-dimensional discrete target trajectory coordinate data is calculated based on the preset interpolation direction gradient vector field, and the inverse vector space bias data is superimposed on the standard two-dimensional discrete target trajectory coordinate data to generate the reconstructed target trajectory coordinates. The reconstructed target trajectory coordinates are mapped to the surface topography feature matrix, and local relative undulation data and local concave-convex curvature feature data are extracted. The Z-axis follow-up height command parameter and the dispensing volume pulse width command parameter are calculated and generated respectively. Based on the reconstructed target trajectory coordinates, the Z-axis follow-up height command parameter, and the dispensing volume pulse width command parameter, the actuator and dispensing valve are controlled to perform a three-dimensional collaborative dispensing operation.

[0015] This invention provides a control system and method for visual painting and adhesive application equipment on metal crafts. It has the following beneficial effects: 1. This invention calculates the inverse vector space offset data corresponding to the standard trajectory coordinates through a preset interpolation direction gradient vector field, and superimposes it with the original coordinates to generate reconstructed target trajectory coordinates. This technical solution can apply position compensation in the opposite direction of the downward slope of the workpiece surface before controlling the actuator's movement, offsetting the fluid slippage distance caused by the paint droplet's own gravity. This ensures that the actual position of the cured paint accurately coincides with the target processing trajectory, solving the problem of easy displacement when applying adhesive to inclined surfaces of metal crafts.

[0016] 2. This invention maps the reconstructed target trajectory coordinates to a surface morphology feature matrix, extracting local relative undulation data and two-dimensional Laplacian operator data characterizing the surface concavity and convexity polarity. The system uses the relative undulation data to calculate the Z-axis follow-up height command, maintaining a constant physical gap between the dispensing valve and the workpiece surface, avoiding paint stringing or dripping caused by sudden gap changes; simultaneously, it uses the Laplacian operator data to calculate the glue volume pulse width command, increasing the glue dispensing pulse width in micro-depression areas and decreasing the glue dispensing pulse width in protrusion areas, achieving adaptive compensation of glue dispensing volume based on the workpiece's micro-morphology, ensuring the uniformity of the final paint coating thickness.

[0017] 3. This invention encapsulates the target pulse counts of the X and Y axes for planar position interpolation, the follow-up height command parameters of the Z axis, and the dispensing volume pulse width command parameters into a unified multi-axis synchronous data packet according to the control protocol. This data packet is then uniformly distributed to the servo driver and dispensing controller via an industrial fieldbus, eliminating the time lag in command distribution between independent hardware communications. This achieves strict synchronization between spatial three-dimensional motion trajectory adjustment and dispensing valve dispensing action, improving the control accuracy and operational stability of electromechanical equipment working collaboratively on complex-shaped workpieces. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the hardware architecture of the control system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the control method flow according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the global baseline parameter acquisition process according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the timing of multimodal light field image acquisition according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the physical morphology feature mapping process according to an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the trajectory reverse reconstruction calculation principle of an embodiment of the present invention; Figure 7This is a schematic diagram illustrating the adaptive coupling calculation principle of spatial height and dispensing volume in an embodiment of the present invention. Figure 8 This is a schematic diagram of the three-dimensional linkage execution process according to an embodiment of the present invention; Figure 9 This is a data distribution diagram of microscopic topography mapping and trajectory reconstruction according to an embodiment of the present invention; Figure 10 This is a data distribution diagram of microscopic topography mapping and trajectory reconstruction according to an embodiment of the present invention.

[0019] Among them, 100 is the control system; 10 is the main control unit; 20 is the vision device; 21 is the area scan camera; 30 is the light source assembly; 31 is the bright field light source; 32 is the dark field light source; 33 is the light source controller; 40 is the actuator; 41 is the X-axis motor; 42 is the Y-axis motor; 43 is the Z-axis motor; 44 is the dispensing valve; and 45 is the dispensing controller. Detailed Implementation

[0020] 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.

[0021] See attached document Figure 1 The present invention provides a control system 100, which includes a main control unit 10, a vision device 20, a light source assembly 30, and an actuator 40.

[0022] The main control unit 10 is electrically connected to the vision device 20, the light source assembly 30 and the actuator 40 respectively, and is used to realize the communication scheduling, timing control and control command generation of each hardware module in the system.

[0023] The vision device 20 includes an area scan camera 21. The area scan camera 21 is mounted on the physical execution end of the device and has a hardware trigger interface that supports external pulse signal input. The area scan camera 21 and the dispensing valve 44 are fixedly mounted on the same physical execution end. There are preset relative installation offsets in the X, Y, and Z directions between the optical axis reference point of the area scan camera 21 and the center of the dispensing port of the dispensing valve 44. The relative installation offsets are obtained after the device assembly calibration and stored in the main control unit 10 for use by the main control unit 10 in coordinate conversion, trajectory compensation, and Z-axis height control.

[0024] The X and Y relative installation offsets are used to convert the visual detection position of the area array camera 21 to the actual dispensing position of the dispensing valve 44, or to convert the target dispensing position of the dispensing valve 44 to the visual observation position corresponding to the area array camera 21; the Z relative installation offset is used to convert the focusing height reference of the area array camera 21 to the working height reference corresponding to the dispensing port of the dispensing valve 44.

[0025] The light source assembly 30 includes a bright field light source 31, a dark field light source 32, and a light source controller 33. The light source controller 33 is connected to the bright field light source 31 and the dark field light source 32 respectively, and is used to control the bright field light source 31 and the dark field light source 32 to perform strobe light emission according to the trigger signal issued by the main control unit 10.

[0026] The actuator 40 includes an X-axis motor 41, a Y-axis motor 42, a Z-axis motor 43, a dispensing valve 44, and a dispensing controller 45. The dispensing controller 45 is electrically connected to the main control unit 10 and the dispensing valve 44, and is used to receive the dispensing volume pulse width command parameter issued by the main control unit 10, and output the PWM signal or high / low level pulse signal corresponding to the command parameter to the dispensing valve 44.

[0027] The dispensing valve 44 is equipped with a pulse width adjustment terminal. The pulse width adjustment terminal of the dispensing valve 44 participates in the underlying interpolation control as a virtual dispensing axis in the actuator 40. It is used to receive the PWM signal or high and low level pulse signal output by the dispensing controller 45. The main control unit 10 sends the glue volume pulse width command parameter to the dispensing controller 45, so that the dispensing controller 45 adjusts the effective level duration of the dispensing valve 44 in a single interpolation cycle, thereby controlling the single dispensing volume of the dispensing valve 44.

[0028] The control method of the present invention includes obtaining a reference height, acquiring multimodal images of the area to be detected, constructing a surface morphology feature matrix, compensating for reverse slippage of a standard trajectory, calculating the Z-axis follow-up height, calculating the dispensing volume pulse width, and executing multiple axes synchronously.

[0029] Specifically, the main control unit 10 first obtains the global absolute reference height parameter of the current workpiece batch through the Z-axis focusing process of the area array camera 21; then, it acquires bright field and dark field images in the area to be detected, and constructs a surface morphology feature matrix based on the gray-scale gradient difference value; subsequently, it performs reverse slip compensation on the standard two-dimensional discrete target trajectory coordinates under a unified coordinate reference, and generates a Z-axis follow-up height command based on local relative undulation data, and generates a dispensing volume pulse width command based on local concave and convex curvature features; finally, the main control unit 10 encapsulates the planar trajectory, Z-axis height and dispensing pulse width into a low-level discrete interpolation command queue according to the same interpolation cycle, and sends it to the execution mechanism 40 to complete the adaptive dispensing operation.

[0030] See attached document Figure 2The present invention provides a control method, comprising the following steps: S110, the main control unit 10 sends a drive command to the actuator 40, controlling the area scan camera 21 to move above the positioning reference feature area; subsequently, the main control unit 10 controls the Z-axis motor 43 to perform step displacement within a preset stroke, and receives the sequence of images acquired by the area scan camera 21 at each step position; the main control unit 10 extracts edge sharpness features from the sequence of images, calculates the sharpness evaluation value corresponding to each image, and determines the physical encoder coordinates of the Z-axis motor 43 corresponding to the maximum sharpness evaluation value; the main control unit 10 saves the physical encoder coordinates as a global absolute reference height parameter; S120, the main control unit 10 sends a displacement command to the actuator 40 according to the coordinates of the area to be detected, controlling the area scan camera 21 to move above the area to be detected. After the area scan camera 21 reaches the acquisition position of the area to be detected, the main control unit 10 sends a coordinate locking command to the actuator 40, causing the X-axis motor 41, Y-axis motor 42, and Z-axis motor 43 to stop moving. While the actuator 40 remains stationary, the main control unit 10 simultaneously sends a high-frequency trigger level sequence to the light source controller 33 and the area scan camera 21. The light source controller 33 controls the bright field light source 31 and the dark field light source 32 to light up sequentially according to the high-frequency trigger level sequence. The area scan camera 21 completes image capture according to the same high-frequency trigger level sequence and outputs the bright field image corresponding to the bright field light source 31 and the dark field image corresponding to the dark field light source 32. S130, the main control unit 10 extracts target pixel region data from the bright field image and the dark field image. The target pixel region data can be determined based on the circumscribed region after mapping the standard two-dimensional discrete target trajectory coordinates through the hand-eye calibration matrix, or it can be determined based on a preset ROI template, the metal craft boundary detection results, or a combination of the above methods. The main control unit 10 performs brightness normalization and filtering preprocessing on the target pixel region data, and calculates the gray-level gradient difference value of the corresponding spatial coordinates. Subsequently, the main control unit 10 performs nonlinear inverse mapping operation based on the gray-level gradient difference value to generate a surface morphology feature matrix used to characterize the relative undulation trend of the metal craft surface.

[0031] In this embodiment, the surface topography feature matrix is ​​used to represent the local relative undulation of the surface of the metal craft relative to a preset flat reference surface, and is not limited to the absolute measurement value of the actual three-dimensional height of the workpiece. After generating the surface topography feature matrix, the main control unit 10 uses the neighborhood mean of the positioning reference feature area, the flat reference area, or the target area as the relative undulation zero reference to perform zero-point normalization processing on the surface topography feature matrix; the positive direction of the surface topography feature matrix is ​​defined as the direction of increasing concavity relative to the zero reference.

[0032] The main control unit 10 further extracts the partial derivative data of the surface morphology feature matrix on a two-dimensional orthogonal plane and constructs an interpolation directional gradient vector field corresponding to the surface morphology feature matrix. The interpolation directional gradient vector field is used to characterize the direction and rate of change of local undulations at the corresponding coordinate point; when the positive direction of the surface morphology feature matrix is ​​defined as the direction of increasing concavity, the direction vector in the interpolation directional gradient vector field is used to characterize the direction of the fastest increase in concavity at the corresponding coordinate point and serves as the basis for estimating the potential slip direction of the paint liquid.

[0033] S140, the main control unit 10 extracts the pre-stored standard two-dimensional discrete target trajectory coordinate data. The standard two-dimensional discrete target trajectory coordinate data is defined as the target glue-dropping coordinates of the glue outlet of the dispensing valve 44 in the mechanical motion coordinate system; the main control unit 10 calls the pre-stored hand-eye calibration matrix, and combines the X and Y relative installation offsets between the area array camera 21 and the dispensing valve 44 to map the standard two-dimensional discrete target trajectory coordinate data defined in the mechanical motion coordinate system to the pixel reference system where the surface morphology feature matrix is ​​located, so as to retrieve the local morphology feature data corresponding to the actual target glue-dropping position in the visual pixel coordinates.

[0034] The main control unit 10 retrieves the local directional gradient vector corresponding to the mapped pixel trajectory coordinates in the interpolation directional gradient vector field, and converts the local directional gradient vector in the pixel reference system into the physical directional gradient vector in the mechanical motion coordinate system according to the scaling and rotation parameters in the hand-eye calibration matrix.

[0035] Subsequently, the main control unit 10 retrieves the fluid slip compensation coefficient and multiplies the physical direction gradient vector in the mechanical motion coordinate system with the fluid slip compensation coefficient to obtain the inverse vector space offset data. When the physical direction gradient vector is normalized to a dimensionless direction vector, the fluid slip compensation coefficient represents the compensation distance along that direction, and its unit is consistent with the standard two-dimensional discrete target trajectory coordinate unit in the mechanical motion coordinate system; when the physical direction gradient vector retains the slope magnitude, the fluid slip compensation coefficient represents the proportional coefficient that maps the local slope to the planar compensation displacement.

[0036] The main control unit 10 performs superposition and translation calculations on the inverse vector space offset data and the standard two-dimensional discrete target trajectory coordinates to generate reconstructed target trajectory coordinates.

[0037] S150, the main control unit 10 extracts the local relative undulation data corresponding to the reconstructed target trajectory coordinates in the surface topography feature matrix. The local relative undulation data is the surface topography feature matrix value after zero-point normalization, used to characterize the local height change trend of the current target glue application position relative to the flat reference area; among them, the global absolute reference height parameter is used to correct the overall height translation error of the current workpiece batch, and the local relative undulation data is used to correct the local micro-undulation error of the same workpiece surface. The main control unit 10 combines the local relative undulation data, the global absolute reference height parameter, and the Z-axis relative installation offset between the area array camera 21 and the dispensing valve 44 to calculate and generate the Z-axis follow-up height command parameter for the current interpolation cycle; the main control unit 10 calculates the two-dimensional Laplacian operator data corresponding to the reconstructed target trajectory coordinates in the surface topography feature matrix, and uses the two-dimensional Laplacian operator data as the local concavity and convexity curvature feature data; subsequently, the main control unit 10 calculates and generates the dispensing volume pulse width command parameter for the current interpolation cycle based on the local concavity and convexity curvature feature data and the reference dispensing pulse width constant; S160, the main control unit 10 integrates the reconstructed target trajectory coordinates, Z-axis follow-up height command parameters, and dispensing volume pulse width command parameters into multi-dimensional collaborative control data, and encapsulates them into a low-level discrete interpolation command queue according to a unified interpolation cycle or a unified timestamp. Each command in the low-level discrete interpolation command queue includes at least the trajectory point number, X-axis target position or target pulse count, Y-axis target position or target pulse count, Z-axis follow-up height command parameters, dispensing volume pulse width command parameters, and the corresponding interpolation cycle identifier or timestamp. The main control unit 10 sends the low-level discrete interpolation command queue to the execution mechanism 40. The execution mechanism 40 parses the commands within the corresponding interpolation cycle and synchronously drives the X-axis motor 41 and Y-axis motor 42 to perform reverse bias trajectory displacement, drives the Z-axis motor 43 to perform follow-up height adjustment, and simultaneously controls the dispensing valve 44 to output an effective high-level duration or duty cycle electrical signal corresponding to the dispensing volume pulse width command parameter.

[0038] See attached document Figure 3 In conjunction with the above method embodiments, it is mainly used to establish an absolute height reference before formally collecting deformation characteristics of metal crafts. The process of the main control unit 10 acquiring the global absolute reference height parameter specifically includes the following: In this embodiment, the main control unit 10 sends a displacement drive control signal to the actuator 40.

[0039] The actuator 40 drives the area array camera 21 to move directly above the positioning reference feature area of ​​the metal craft according to the drive control signal. The positioning reference feature area can be a pre-defined area with edge contour texture or a silkscreen mark point on the surface of the metal craft.

[0040] The main control unit 10 determines a range of travel that fluctuates up and down based on the preset system nominal focus height.

[0041] As an optional implementation, the nominal focusing height of the system is typically determined based on the mechanical focusing zero point during equipment assembly and the theoretical thickness of the metalwork drawing. The range of the upward and downward floating travel interval is typically set to 1.5 to 2 times the theoretical thickness tolerance, while the displacement step size is set according to the depth-of-field parameters of the area scan camera 21 lens, for example, set to one-quarter or one-third of the depth of field, to ensure that sufficiently dense image samples can be acquired near the focal plane. The main control unit 10 sends a stepping motion command to the Z-axis motor 43, controlling the Z-axis motor 43 to drive the area scan camera 21 to perform discrete stepping displacement operations along the optical axis within the travel interval according to the set displacement step size. During the stop interval of each stepping displacement, the main control unit 10 triggers the area scan camera 21 to acquire an image frame of the current position. As the Z-axis motor 43 completes the stepping operation of the entire travel interval, the area scan camera 21 outputs a sequence of images captured at different physical heights.

[0042] After receiving the sequence of images, the main control unit 10 extracts the image blocks corresponding to the positioning reference feature area in the sequence of images, performs edge sharpness feature extraction, and calculates the sharpness evaluation value.

[0043] From the perspective of physical optical imaging principles, the closer the distance between the lens of the area array camera 21 and the workpiece surface is to the ideal focal length, the stronger the contrast of texture edges in the resulting image, i.e., the more drastic the grayscale changes. Based on this characteristic, the main control unit 10 uses a sharpness evaluation function to quantify the focus degree of each frame of the image. In specific implementation, the main control unit 10 uses the Tenengrad function model to calculate the sharpness evaluation value, and the calculation formula is as follows: ; In the formula, This indicates that the Z-axis motor 43 is located at the physical height. The sharpness evaluation value of the corresponding image frame at that time; and Represents the orthogonal coordinates of pixels within an image block; Represents pixels Edge gradient values ​​along the horizontal direction of the image; Represents pixels Edge gradient values ​​along the vertical direction of the image.

[0044] The edge gradient values ​​in the horizontal and vertical directions of an image can be obtained using the Sobel operator, the Prewitt operator, or other conventional gradient operators.

[0045] The main control unit 10 compares all the sharpness evaluation values ​​corresponding to the sequence images, determines the maximum sharpness evaluation value, and extracts the physical encoder coordinates fed back by the internal encoder of the Z-axis motor 43 when the maximum sharpness evaluation value is generated.

[0046] To further improve the accuracy of height reference determination, the main control unit 10 retrieves the maximum sharpness evaluation value and the sharpness evaluation values ​​of two adjacent step positions in the stroke sequence, performs parabolic extreme value fitting calculation, and uses the continuous domain coordinates corresponding to the vertex of the parabola as the corrected physical encoder coordinates.

[0047] In actual operation, if the maximum sharpness evaluation value is located at the end point of the travel interval, the main control unit 10 determines that the current scanning travel does not cover the real focal plane. At this time, the main control unit 10 takes the end point as the new scanning starting point, expands the travel interval in the corresponding direction, and re-executes the step acquisition and sharpness evaluation process until the maximum sharpness evaluation value is located within the effective travel interval.

[0048] The main control unit 10 saves the corrected physical encoder coordinates as a global absolute reference height parameter. Global absolute reference height parameter It is used to characterize the absolute height reference of the metal crafts in the current batch of processing in the camera detection coordinate system, and to correct the overall height translation caused by material batch differences, workpiece thickness deviations or loading / unloading assembly errors.

[0049] Since the optimal focusing height of the area scan camera 21 is not exactly the same as the process working height of the dispensing valve 44, when the main control unit 10 generates the Z-axis follow-up height command for the dispensing valve 44, it calls the relative installation offset in the Z direction between the area scan camera 21 and the dispensing valve 44, and combines it with the preset dispensing working gap to set the global absolute reference height parameter. This is converted to the Z-axis working height reference corresponding to the dispensing port of dispensing valve 44.

[0050] See attached document Figure 4 With the actuator remaining stationary, optical response data of the metal craft surface under different incident light fields are acquired. The process of high-frequency timing acquisition of the multi-modal light field controlled by the main control unit 10 is as follows: In this embodiment, after the main control unit 10 completes the acquisition of absolute reference parameters and determines the detection area, it controls the actuator 40 to move the area scan camera 21 above the detection area according to the detection area coordinates. When the detection area is divided into multiple acquisition sub-areas, the main control unit 10 controls the area scan camera 21 to move sequentially to the acquisition position corresponding to each acquisition sub-area.

[0051] After the area scan camera 21 reaches the corresponding acquisition position, the main control unit 10 sends a coordinate locking command to the actuator 40. Upon receiving the coordinate locking command, the X-axis motor 41, Y-axis motor 42, and Z-axis motor 43 stop their displacement operations and enter a position holding state under the control closed loop of the servo driver. This reduces mechanical micro-movements during continuous acquisition of bright-field and dark-field images, ensuring that different light field images maintain consistent physical spatial positions.

[0052] After the actuator 40 is in the position holding state, the main control unit 10 synchronously sends a high-frequency trigger level sequence to the light source controller 33 and the area scan camera 21 via a hardware interface. The high-frequency trigger level sequence includes a first pulse and a second pulse arranged sequentially along the time axis. When generating the high-frequency trigger level sequence, the main control unit 10 ensures that the time interval between the first pulse and the second pulse is greater than the sum of the single exposure time and the internal data readout time of the area scan camera 21, in order to avoid overlap of the exposure windows or readout processes of two adjacent frames. The time interval can be set to milliseconds or sub-milliseconds according to the exposure parameters and readout rate of the area scan camera 21.

[0053] After receiving the high-frequency trigger level sequence, the light source controller 33 performs strobe control based on the edge changes or level states of the pulse signal. As an optional implementation, during the effective level of the first pulse, the light source controller 33 drives the bright-field light source 31 to emit light. The bright-field light source 31 provides an approximately perpendicularly incident illumination beam to the surface of the metal craft. The area array camera 21 completes exposure within the trigger window corresponding to the first pulse, acquires reflection information under this illumination state, and generates a bright-field image. .in, and Represents pixel coordinates within the image plane. Bright-field image. It is mainly used to reflect the basic reflective texture, outline boundary and two-dimensional positioning features of the surface of metal crafts.

[0054] During the effective level of the second pulse, the light source controller 33 controls the bright field light source 31 to turn off and drives the dark field light source 32 to emit light. The dark field light source 32 provides a low-angle, obliquely incident illumination beam to the surface of the metal craft. As an optional implementation, the low angle is set to 10 to 30 degrees to reduce the direct specular reflection effect of the highly reflective metal surface on the camera. The area scan camera 21 re-exposes within the trigger window corresponding to the second pulse to generate a dark field image. .

[0055] When the surface of a metal craft has minute warps, depressions, or protrusions, the scattering response of low-angle oblique incident light in these areas will change, thus affecting the dark field image. It can enhance the grayscale contrast in micro-deformation areas. The main control unit 10 receives the bright-field image transmitted by the area array camera 21. and dark field images The two data points are stored in the system cache for subsequent grayscale gradient difference and surface morphology feature calculations.

[0056] The area scan camera 21 can use the image acquisition link of a conventional industrial camera to convert the analog electrical signal generated by the light receiving element into a digital image matrix, and transmit it to the main control unit 10 through gigabit network, USB 3.0, CameraLink or other industrial vision communication interfaces.

[0057] When the area to be processed of the metal craft or the coverage of the standard two-dimensional discrete target trajectory coordinates is larger than the single imaging field of view of the area array camera 21, the main control unit 10 divides the area to be detected into multiple acquisition sub-regions with overlapping boundaries according to the circumscribed rectangular area of ​​the target trajectory or the preset detection area. The main control unit 10 controls the X-axis motor 41 and the Y-axis motor 42 to move sequentially to the positions corresponding to each acquisition sub-region, and performs bright field image and dark field image acquisition in each acquisition sub-region respectively.

[0058] The main control unit 10 fuses the surface topography feature matrices of multiple acquisition sub-regions into the same global coordinate system based on the machine coordinates, hand-eye calibration matrix, and registration results of overlapping areas of the images corresponding to each acquisition sub-region. Specifically, the main control unit 10 first determines the initial position of the corresponding surface topography feature matrix in the global coordinate system based on the machine coordinates of each acquisition sub-region, and then performs registration correction using the overlapping areas of adjacent acquisition sub-regions. For topography data repeatedly obtained in the overlapping areas, the main control unit 10 can use weighted averaging, confidence-first, or median fusion methods to synthesize the data, thereby generating a global surface topography feature matrix covering the area to be detected.

[0059] See attached document Figure 5 The two-dimensional image data acquired through multimodal light field acquisition is converted into morphological feature data characterizing the relative undulations and local variation trends of the surface of metal crafts. The process of the main control unit 10 constructing the surface morphological feature matrix and the interpolation direction gradient vector field is as follows: In this embodiment, the main control unit 10 reads the bright-field image from the system cache. With dark field images Considering that the reflective properties of metal crafts surfaces can easily cause local speckle noise, the main control unit 10 first performs smoothing preprocessing on the bright field image and the dark field image respectively; as one implementation method, the smoothing preprocessing can use a Gaussian filtering algorithm.

[0060] Before or after smoothing preprocessing, the main control unit 10 can also perform exposure normalization, brightness normalization or reflectance normalization on bright field images and dark field images to reduce the impact of surface reflectance differences of different batches of metal crafts, light source brightness fluctuations and local high reflectance areas on subsequent grayscale gradient difference calculations.

[0061] The main control unit 10 extracts target pixel region data corresponding to the same cutting boundary from the preprocessed bright-field image and dark-field image. After the above preprocessing, the macroscopic contour and microscopic deformation edge features in the image are preserved, and the interference of high-frequency noise on subsequent gradient calculation and difference operation is reduced accordingly.

[0062] The main control unit 10 calculates the pixel gradients of corresponding spatial coordinates for the preprocessed bright-field and dark-field images, and performs difference operations to obtain grayscale gradient difference values. As an optional implementation, the main control unit 10 calculates coordinate points... Gray gradient difference at the location The computational model is configured as follows: ; In the formula, Represents the gradient operator; Indicates the brightfield image at coordinate points The local gradient vector at that point; Indicates the dark field image at coordinate point The local gradient vector at that point.

[0063] Because bright-field beams primarily use an approximately perpendicular incidence method, while dark-field beams primarily use a low-angle oblique incidence method, their optical responses to minute undulations on the metal surface differ. Therefore, the gray-scale gradient difference value... It can weaken the influence of the intrinsic reflectivity of the metal surface to a certain extent, so that the difference results can better reflect the scattering changes caused by local micro curvature perturbations.

[0064] The main control unit 10 performs a nonlinear inverse mapping operation based on the gray-level gradient difference value to generate a surface morphology feature matrix corresponding to the surface morphology features of the metal craft. The surface morphology feature matrix is ​​used to characterize the relative undulation trend and local unevenness of the surface of the metal craft, and is not limited to the absolute measurement value of the actual three-dimensional height of the workpiece.

[0065] Since the optical scattering response and surface deformation slope are usually not strictly linearly related, the master control unit 10 can use a polynomial model based on the evolution of the bidirectional reflection distribution function for inverse derivation. As one implementation method, the elements in the surface topography feature matrix... The specific solution formula is set as follows: ; In the formula, Represents coordinate points The value at which the relative undulation deformation is represented; The magnitude of the gray-level gradient difference vector is represented by the Euclidean norm of the difference between the two gradient vectors at the corresponding coordinate points. This is the empirical coefficient for a single reflection. This is the empirical coefficient for secondary reflection.

[0066] First reflection empirical coefficient With the empirical coefficient of secondary reflection This information can be obtained during the equipment's factory calibration phase. Specifically, the main control unit 10 acquires image data of a standard block with known roughness, relative height, tilt angle, or curvature under bright-field light source 31 and dark-field light source 32. It calculates the gray-level gradient difference values ​​at different known morphological locations and uses the known relative height, tilt angle, or curvature of the corresponding position of the standard block as the fitting target. Through the least squares method or other fitting algorithms, it establishes a mapping relationship between the gray-level gradient difference values ​​and the relative morphological feature values, thereby obtaining the empirical coefficient of primary reflection. and empirical coefficient of secondary reflection .

[0067] The main control unit 10 traverses each spatial coordinate point within the target pixel region, calculates the relative deformation value corresponding to each coordinate point, thereby forming a discrete two-dimensional surface morphology feature matrix. This surface morphology feature matrix realizes the numerical expression of the small deformation trend of the metal surface and provides a data foundation for subsequent trajectory offset, Z-axis height compensation, and adhesive dispensing pulse width compensation.

[0068] Subsequently, the main control unit 10 extracts the partial derivative data of the surface topography feature matrix in the two-dimensional orthogonal plane, and constructs the interpolation direction gradient vector field corresponding to the surface topography feature matrix. In actual calculation, the main control unit 10 can perform central difference operations on the surface topography feature matrix in the X-axis and Y-axis directions respectively to approximate continuous partial derivatives and generate local partial derivatives. The difference operation formula can be expressed as: ; ; In the formula, and These represent the partial derivative values ​​of the surface topography feature matrix in the horizontal X-direction and the vertical Y-direction, respectively. For the coordinate points on the boundary of the surface topography feature matrix, since there is a lack of complete adjacent data, the main control unit 10 uses forward difference or backward difference to replace the central difference for partial derivative calculation to avoid computational gaps at the boundary positions.

[0069] The main control unit 10 will set the coordinate points The corresponding partial derivatives in two directions are combined into a column vector to generate a local gradient vector. , is represented as: ; The main control unit 10 aggregates the local directional gradient vectors of all coordinate points to form an interpolated directional gradient vector field. The local directional gradient vectors in the vector field... This direction is used to characterize the direction of the fastest relative change in undulation on the surface of a metal craft at that coordinate point; when the positive direction of the surface topography feature matrix is ​​defined as the direction of increasing concavity, this direction can characterize the direction of the fastest increase in concavity and serve as an estimated direction for the potential slip direction of the paint. Local directional gradient vector. The modulus is used to characterize the local tilt rate at the corresponding position.

[0070] Through the above processing, the main control unit 10 converts the luminance difference in the bright field image and the dark field image into topographic gradient data that can be used for geometric offset calculation.

[0071] In this embodiment, the main control unit 10 pre-stores the calibration relationship between the mechanical motion coordinate system, the visual pixel coordinate system, and the dispensing valve outlet coordinate system. The mechanical motion coordinate system is used to represent the physical motion positions of the X-axis motor 41, the Y-axis motor 42, and the Z-axis motor 43; the visual pixel coordinate system is used to represent the pixel positions in the images acquired by the area array camera 21; and the dispensing valve outlet coordinate system is used to represent the actual dispensing point position of the dispensing valve 44.

[0072] The main control unit 10 completes the mutual conversion between the mechanical motion coordinate system and the visual pixel coordinate system through the hand-eye calibration matrix, and completes the conversion between the visual detection position and the actual dispensing position through the relative installation offset between the area scan camera 21 and the dispensing valve 44. Therefore, when performing trajectory compensation, Z-axis height compensation, and dispensing pulse width compensation in the subsequent process, the visual feature data, the actual dispensing position of the dispensing valve, and the physical motion commands of the actuator 40 can all be under the same coordinate reference.

[0073] See attached document Figure 6 In conjunction with the aforementioned method embodiments, this mainly addresses the slippage phenomenon of paint on a small inclined plane caused by physical factors such as gravity. The control trajectory is reconstructed by performing reverse spatial compensation on the dispensing position in advance. The specific process by which the main control unit 10 generates the coordinates of the reconstructed target trajectory includes the following: In this embodiment, the main control unit 10 extracts standard two-dimensional discrete target trajectory coordinate data from the system storage module. This data typically originates from the machining code generated after parsing the product's CAD drawings, and consists of a set of discrete points defined in a mechanical motion coordinate system. Composition, in which This represents the sequence number of the discrete trajectory point. To ensure that the mechanical physical coordinates match the visual pixel coordinates, the main control unit 10 retrieves the pre-stored hand-eye calibration matrix and maps the standard two-dimensional discrete target trajectory coordinate data to the pixel reference system where the aforementioned interpolated directional gradient vector field is located. The hand-eye calibration matrix includes at least the translation parameters, rotation parameters, and scaling parameters between the mechanical motion coordinate system and the visual pixel coordinate system. After completing the local directional gradient vector retrieval in the pixel reference system, the main control unit 10 converts the directional gradient vector in the pixel reference system into a physical directional gradient vector in the mechanical motion coordinate system based on the scaling and rotation parameters of the hand-eye calibration matrix, thus avoiding direct superposition of pixel coordinate data and mechanical coordinate data.

[0074] For the calculation of affine transformation between the mechanical motion coordinate system and the visual pixel coordinate system, those skilled in the art can use the conventional nine-point calibration method to solve the hand-eye calibration matrix. The spatial coordinate transformation technology is a well-known technology in this field and will not be described in detail here.

[0075] The main control unit 10 retrieves the local orientation gradient vector corresponding to the standard two-dimensional discrete target trajectory coordinates in the interpolated orientation gradient vector field. Considering that the trajectory coordinates after hand-eye matrix mapping are mostly floating-point numbers and cannot be directly mapped to integer indices of the pixel matrix, the main control unit 10 adopts a bilinear interpolation algorithm to avoid step-like jitter errors in trajectory calculation caused by direct rounding. In actual processing, if the mapped trajectory coordinates are close to the effective boundary of the interpolation direction gradient vector field, resulting in fewer than four available adjacent integer pixels, the main control unit 10 determines that the coordinates are a boundary point and stops performing bilinear interpolation. Instead, it uses the nearest neighbor interpolation algorithm to directly extract the gradient vector of the single integer pixel closest to it in space as the local direction gradient vector. After completing the nearest neighbor interpolation, the main control unit 10 still converts the local direction gradient vector into a physical direction gradient vector in the mechanical motion coordinate system according to the hand-eye calibration matrix. This processing mechanism effectively avoids logical interruptions caused by boundary calculations exceeding limits.

[0076] Based on fluid dynamics, when the paint extruded from the dispensing valve comes into contact with a product surface with a slight inclination, it typically undergoes physical slippage along the direction of maximum downward inclination under the combined effects of gravity and surface tension. To reduce or counteract this slippage deviation, the main control unit 10 retrieves the fluid slippage compensation coefficient set by the system. The fluid slip compensation coefficient is an empirical constant that comprehensively characterizes physical properties such as the kinematic viscosity of the paint, dispensing height, and curing time. As an optional implementation, the fluid slip compensation coefficient is set to a value between 0.1 and 0.5, and its specific value can be obtained through a stepped inclined spray test in the early stages of the process. The main control unit 10 multiplies the physical direction gradient vector in the mechanical motion coordinate system with the fluid slip compensation coefficient to generate inverse vector space bias data. The calculation formula is expressed as follows: ; In the formula, the negative sign indicates that the compensation direction is strictly opposite to the direction of the maximum downward slope; It is a two-dimensional vector containing offset components corresponding to the X and Y axes. The offset components of the two-dimensional vector are defined in the mechanical motion coordinate system, and its units are consistent with the units of standard two-dimensional discrete target trajectory coordinate data.

[0077] S144, after acquiring the offset component, the main control unit 10 will use the inverse vector space offset data in the mechanical motion coordinate system. Compared with the original standard two-dimensional discrete target trajectory coordinates Superposition and translation calculations are performed. Through vector addition operations, the main control unit 10 generates the reconstructed target trajectory coordinates defined in the mechanical motion coordinate system. The calculation formula is configured as follows: ; Generated reconstructed target trajectory coordinates Essentially, the original planned glue-dropping points on a flat theoretical plane were effectively vector-driven towards the higher microscopic protrusions along the deformed surface. This forward-looking spatial offset calculation helps ensure that the final fluid static solidification position of the paint liquid after gravitational slippage is closer to the preset target contour boundary.

[0078] See attached document Figure 7 In conjunction with the aforementioned method embodiments, the main approach is to map visual features to the physical execution domain, thereby achieving the dynamic conformation of the actuator in three-dimensional space and the dynamic compensation of fluid volume. The specific process by which the main control unit 10 calculates multi-dimensional instruction parameters includes the following: In this embodiment, the main control unit 10 obtains the reconstructed target trajectory coordinates defined in the mechanical motion coordinate system. Then, the hand-eye calibration matrix is ​​invoked to map the reconstructed target trajectory coordinates to the pixel reference system where the surface topography feature matrix is ​​located, and the mapped pixel coordinates are used as indices to look up the corresponding local relative undulation data in the surface topography feature matrix. Since the reconstructed target trajectory coordinates usually contain sub-pixel level floating-point decimals after vector translation, as a preferred implementation, the main control unit 10 uses a bilinear interpolation algorithm to extract the integer pixel values ​​around the coordinates, perform weighted summation, and calculate the corresponding local relative undulation data. If the coordinates are close to the effective boundary of the matrix, resulting in insufficient available adjacent pixels, the main control unit 10 switches to the nearest neighbor interpolation algorithm to extract data, in order to maintain the continuity of the operation process and avoid addressing errors.

[0079] The main control unit 10 will process the local relative fluctuation data. Multiply by the preset height mapping coefficient Converted to physical height offset suitable for the actuator The height mapping coefficient To scale the relative undulation values ​​in the surface topography feature matrix to actual millimeter-level mechanical displacements, those skilled in the art can obtain the specific values ​​through a visual offline calibration program for stepped gauge blocks. Specifically, the main control unit 10 acquires images of stepped gauge blocks with known height differences, calculates the surface topography feature matrix values ​​corresponding to each stepped region, and uses the known height differences as the fitting target to obtain the height mapping coefficients through linear fitting or least squares fitting. Subsequently, the main control unit 10 will calculate the physical height offset. The global absolute reference height parameter obtained in the preceding steps Perform linear superposition calculations to generate the Z-axis follower height command parameter for the current interpolation cycle. Because there is a relative Z-axis installation offset between the area array camera 21 and the dispensing valve 44, and a preset dispensing working gap needs to be maintained between the dispensing outlet of the dispensing valve 44 and the surface of the metal craft, the main control unit 10 further superimposes or subtracts the relative Z-axis installation offset and the preset dispensing working gap when generating the Z-axis follow-up height command parameter. The superposition calculation formula is configured as follows: ; Alternatively, the main control unit 10 according to , , The relative installation offset and the preset dispensing working gap are used together to generate the result. ,in, Used to convert the relative fluctuation values ​​in the surface topography feature matrix into millimeter-level values ​​that can be executed by the actuator 40. Displacement compensation amount; The relative installation offset is used to compensate for the assembly difference between the focus height reference of the area array camera 21 and the dispensing port height reference of the dispensing valve 44; the preset dispensing working gap is used to maintain a stable distance between the dispensing port of the dispensing valve 44 and the surface of the metal craftwork as required by the process.

[0080] To prevent mechanical collisions caused by excessive calculated height offset due to localized abnormal reflections, the main control unit 10 limits the Z-axis follow-up height command parameter, ensuring its value remains within the safe soft limit range of the Z-axis motor 43. This safe soft limit range is typically preset by the mechanical travel boundary during initial equipment assembly and the parameters of the underlying servo driver.

[0081] To address the issue of surface volume loss or redundancy caused by microscopic deformation of metals, the main control unit 10 calculates the two-dimensional Laplacian operator data corresponding to the reconstructed target trajectory coordinates in the surface topography feature matrix as local concavity / convexity curvature feature data. The Laplacian operator is used to describe the divergence characteristics of spatial surfaces, objectively characterizing the local concavity / convexity polarity and steepness of the workpiece surface. As an optional implementation, the main control unit 10 uses a discrete difference template for calculation, with the following formula: ; In the formula, This represents the Laplace divergence value at the reconstructed target trajectory coordinates; This represents the element value at the index coordinates corresponding to the surface topography feature matrix. In this embodiment, the surface topography feature matrix... The positive direction of the value is defined as the direction of increasing concavity relative to the reference surface. To ensure that the adhesive extrusion compensation direction is consistent with the concavity polarity, the main control unit 10 defines the local concavity curvature feature data L as follows: L is positive when it represents a micro-concave region, and L is negative when it represents a micro-convex region. If the sign direction obtained by directly calculating using the two-dimensional Laplacian operator is opposite to the above definition, the main control unit 10 takes the inverse of the result of the two-dimensional Laplacian operator and uses it as the local concavity curvature feature data L. In other embodiments, the surface topography feature matrix is... The positive direction of the numerical value is defined as the direction of increasing convexity relative to the reference surface. The main control unit 10 then adjusts the correspondence between the local concave-convex curvature feature data and the direction of correction for the dispensing volume pulse width. For the discrete coordinates of the matrix boundary, the main control unit 10 uses a one-sided edge difference method to replace the aforementioned all-directional template to eliminate algorithm dead zones.

[0082] The main control unit 10 retrieves the reference dispensing pulse width constant required for the dispensing valve 44 to operate. This constant represents the duty cycle duration of the bottom layer level required to achieve the standard process adhesive width on a flat surface, and is typically set during equipment commissioning. The main control unit 10 combines two-dimensional Laplacian operator data as local concavity / convexity curvature feature data with volume compensation gain. Calculate and generate the dispensing volume pulse width command parameter for the current interpolation cycle. The computational model is configured as follows: ; In the formula, the volume compensation gain This is the set proportional constant. For this volumetric compensation gain... To obtain the empirical value, those skilled in the art can calibrate it by performing step titration experiments on standard gauge blocks with different divergence characteristics, based on the rheological properties of the paint and the dispensing valve 44's outlet diameter. Specifically, volume compensation gain This is used to convert local concave-convex curvature feature data into the effective level duration increment of the dispensing valve 44, with the unit determined according to the dimensions of the local concave-convex curvature feature data; when the local concave-convex curvature feature data is a dimensionless value... The unit is time. The aforementioned calculation mechanism causes the dispensing valve 44 to extend the pulse width when passing through micro-depression areas to increase paint filling volume, and to correspondingly shorten the pulse width when passing through micro-protrusion areas to reduce adhesive output. To avoid underlying hardware driver malfunctions, the main control unit 10 controls the dispensing volume pulse width command parameter. Upper and lower limit amplitude constraints are applied. The main control unit 10 will truncate instructions that exceed the safe range, which helps to ensure that the final output pulse width value is not lower than the minimum mechanical response time of the dispensing valve 44 and not greater than the maximum allowable cycle time of the current system interpolation cycle.

[0083] To improve the stability of the control process, before generating the Z-axis follow-up height command parameter and the dispensing volume pulse width command parameter, the main control unit 10 performs a validity check on the surface topography feature matrix, the interpolation direction gradient vector field, and the two-dimensional Laplacian operator data. When the gray value at the target coordinate is in the saturation range, the gray-level gradient difference exceeds a preset anomaly threshold, the magnitude of the local direction gradient vector exceeds a preset gradient threshold, or the two-dimensional Laplacian operator data exceeds a preset divergence threshold, the main control unit 10 determines that the location is an abnormal reflection point or an invalid detection point, and replaces it with the neighborhood mean, median filtering result, or data from the previous effective interpolation cycle. As a preferred method, the main control unit 10 prioritizes using the neighborhood median filtering result to replace invalid detection points; when the number of effective pixels in the neighborhood is insufficient, the neighborhood mean is used as a replacement; when effective data still cannot be obtained in the neighborhood, the topography feature data, the Z-axis follow-up height command parameter, or the dispensing volume pulse width command parameter corresponding to the previous effective interpolation cycle are used as replacements.

[0084] See attached document Figure 8 In conjunction with the aforementioned method embodiments, this method is mainly used to convert the logical operation data generated by the main control unit 10 into electrical signals executable by the underlying hardware, thereby completing the final adaptive physical spraying operation. The specific process by which the main control unit 10 controls the actuator 40 and the dispensing valve 44 to perform multi-dimensional collaborative execution includes the following: In this embodiment, the main control unit 10 extracts the reconstructed target trajectory coordinates generated in the aforementioned steps. The number of X-axis target pulses is converted into a basic motion pulse count recognizable by the X-axis motor 41 and Y-axis motor 42 by dividing it by the set mechanical pulse equivalent constant. Since the reconstructed target trajectory coordinates contain floating-point numbers, direct division by the pulse equivalent often produces decimal pulses, leading to errors in reading the underlying registers. To eliminate the unidirectional cumulative error caused by sub-pulse rounding and avoid coordinate drift during long-path execution, as a preferred implementation, the main control unit 10 incorporates rounding calculations during the conversion process. Target pulse number along the Y-axis The calculation formula is configured as follows: ; ; In the formula, and These represent the mechanical pulse equivalents of the X and Y axes, respectively, which are the physical displacements corresponding to a single pulse of motor rotation. Their values ​​are usually determined by the lead of the bottom-level transmission screw and the resolution of the motor servo encoder. This indicates rounding, used to accurately convert floating-point coordinates into the closest valid physical pulse.

[0085] The main control unit 10 will calculate the basic motion pulse count and the Z-axis follow-up height command parameter. and dispensing volume pulse width command parameters The data is encapsulated into a multi-axis synchronous data packet according to the set motion control protocol. The main control unit 10 sends the multi-axis synchronous data packet to the underlying multi-axis servo drive and the independent dispensing controller via the industrial fieldbus.

[0086] For communication transmission of industrial fieldbus and control of the underlying current loop, speed loop and position loop of servo drive, those skilled in the art can configure conventional EtherCAT or CANopen bus protocols and call standard servo algorithms. The communication and underlying servo control technology of motors are well known in the field and will not be described in detail here.

[0087] After receiving the command, the multi-axis servo driver synchronously parses the multi-axis synchronization data packet according to a unified interpolation cycle or a unified timestamp. During actual physical execution, the multi-axis servo driver drives the X-axis motor 41 and Y-axis motor 42 to perform planar position interpolation based on the basic motion pulse count, causing the dispensing valve 44 to move along the anti-slip compensated path. Simultaneously, the multi-axis servo driver drives the Z-axis motor 43 according to the Z-axis follow-up height command parameter. By performing micro-motion tracking in the vertical direction, a constant spatial physical gap is maintained between the dispensing port of the dispensing valve 44 and the surface of the micro-undulating metal craft, which helps to reduce or prevent fluid stringing or dripping caused by gap fluctuations.

[0088] During the trigger window period when the actuator 40 moves the dispensing valve 44 to the discrete node position, the dispensing controller 45 synchronously resolves the... The corresponding effective high-level duration or duty cycle electrical signal triggers a window period that corresponds to the trajectory point number of the current interpolation cycle, establishing a correspondence between the planar displacement of the X-axis motor 41 and Y-axis motor 42, the follow-up height adjustment of the Z-axis motor 43, and the effective dispensing pulse width of the dispensing valve 44 under the same trajectory point number. The dispensing valve 44 responds with an extended pulse width in micro-depression areas to spray more paint for filling deep pits, and a shortened pulse width in micro-protrusion areas to reduce paint discharge. Through the multi-dimensional dynamic coupling adjustment of spatial position anti-slip bias, Z-axis height follow-up, and dispensing volume adaptive compensation, the paint, after contacting the metal surface and undergoing fluid static curing, facilitates the formation of a coating layer with a more consistent contour and relatively uniform thickness on irregular craft surfaces. After all target trajectory points have been executed, the main control unit 10 sends a reset command, controlling the actuator 40 to return to the mechanical origin and enter the next processing standby state.

[0089] Before issuing the underlying discrete interpolation command queue, the main control unit 10 performs safety constraint judgments on the control parameters corresponding to the X-axis motor 41, Y-axis motor 42, Z-axis motor 43, and dispensing valve 44. Specifically, the main control unit 10 determines whether the reconstructed target trajectory coordinates are within the mechanical soft limit range of the X and Y axes, whether the Z-axis follow-up height command parameter is within the Z-axis safety soft limit range, and whether the dispensing volume pulse width command parameter is not less than the minimum mechanical response time of the dispensing valve 44 and not greater than the maximum effective pulse width allowed in the current interpolation cycle. When any control parameter exceeds the corresponding safety range, the main control unit 10 performs truncation, alarm, or suspension of execution for that control parameter.

[0090] To further aid in understanding the adaptive compensation technology solution under multimodal light field vision guidance of this invention, a specific application example is given below, along with corresponding experimental verification and effect comparison data, in conjunction with a specific industrial application scenario and the corresponding numerical calculation process.

[0091] See attached document Figure 9 This figure objectively reflects the spatial coordinate changes of a discrete trajectory point before and after the introduction of visual adaptive compensation.

[0092] In this application embodiment, the object to be processed is a copper commemorative medal with a slightly warped surface from stamping. The process requires applying a layer of transparent enamel paint within a specific contour on its surface for protection. The key parameters preset by the system are as follows: Global absolute reference height parameter. mm; empirical coefficient for primary reflection Secondary reflection empirical coefficient Fluid slip compensation coefficient ; height mapping coefficient mm; reference glue pulse width constant ms, volume compensation gain ms; mechanical pulse equivalents along the X and Y axes mm / pulse.

[0093] After parsing the product's CAD drawing, the main control unit 10 extracts the coordinates of a standard two-dimensional discrete target trajectory point on the contour boundary. In the light field image acquisition and shape feature mapping stage, the main control unit 10 obtains the magnitude of the gray-level gradient difference vector by calculating the gradients of the bright field and dark field images corresponding to the coordinate point. Substitute the values ​​into the polynomial model to calculate the relative undulation deformation at that point. After central difference calculation, the main control unit 10 obtains the partial derivative values ​​of the pixels surrounding the point and generates a local directional gradient vector. This vector physically represents the existence of a microscopic downslope at this location, tilted towards the positive X direction and the negative Y direction.

[0094] Entering the trajectory reverse reconstruction calculation stage, the main control unit 10 calculates the reverse vector space bias data. mm. The generated reconstructed target trajectory coordinates after superposition and translation. mm. Through this calculation, the system reverses the flow of adhesive towards the higher part of the slope to allow space for fluid slippage.

[0095] During the adaptive coupling stage between spatial height and dispensing volume, the main control unit 10 calculates the Z-axis follow-up height command parameter. The Z-axis is slightly raised by 0.18 mm to maintain a safe clearance. Simultaneously, the main control unit 10 calculates the Laplace divergence value around this point. This negative value indicates that this area is a microscopic protrusion. Substituting it into the volumetric calculation model, the ejector volume pulse width command parameter is obtained. The system correspondingly reduced the glue spraying time by 1.0ms to prevent excessive local paint buildup.

[0096] To verify the actual technical effect of the above technical solution, this application provides further experimental verification data.

[0097] See attached document Figure 10 This experiment selected 100 copper commemorative medals with microscopic deformation formed by stamping in the same batch as test samples, and randomly divided them into a control group (50 pieces) and an experimental group (50 pieces). The control group adopted the traditional planar two-dimensional dispensing trajectory with fixed height and fixed pulse width control strategy; the experimental group adopted the adaptive compensation control strategy under multimodal light field vision guidance provided by this invention.

[0098] After the experiment, a high-precision laser confocal microscope was used to perform three-dimensional contour scanning of the cured paint coatings of the two groups of samples. The core evaluation indicators were set as "edge overflow rate" (i.e., the proportion of trajectory points where the paint crosses the set boundary area) and "thickness standard deviation" (characterizing the uniformity of the overall thickness of the coating).

[0099] Statistical data shows that in the control group, due to the lack of intervention for the microscopic tilting and undulations caused by stamping stress, the paint liquid underwent disordered slippage under gravity, resulting in an average edge overflow rate of 14.5% and a coating thickness standard deviation as high as 24.3 μm. The product surface exhibited obvious localized glue buildup and bottom-penetration defects. In the experimental group using the proposed solution, the average edge overflow rate was significantly reduced to 1.2%, with most of the paint liquid precisely adhering to the target contour edge after slippage thanks to the offset points reconstructed through reverse engineering. Simultaneously, the coating thickness standard deviation was reduced to 5.8 μm.

[0100] Combined with appendix Figure 10 The scatter plot curves visually show that the experimental group (attached) Figure 10 The thickness error distribution in the solid line / densely dotted area clearly converges to near zero, with minimal fluctuation; while the control group (attached) Figure 10 The error data (in the dashed / discrete scatter area) exhibits a large-scale divergence. The above experimental data objectively demonstrate that this invention, through optical feature mapping, trajectory inverse reconstruction, and multi-dimensional coupling compensation of height and volume, effectively overcomes the interference of minute deformations on the surface of highly reflective metal workpieces on fluid processes, and significantly improves the spray coating quality on the surface of the craft.

[0101] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control system for a visual painting and adhesive application equipment for metal crafts, characterized in that, include: The data extraction module is used to acquire standard two-dimensional discrete target trajectory coordinate data and surface morphology feature data of the metal crafts to be processed; The trajectory reconstruction calculation module is used to calculate the inverse vector space bias data corresponding to the standard two-dimensional discrete target trajectory coordinate data based on the preset interpolation direction gradient vector field, and to superimpose the inverse vector space bias data with the standard two-dimensional discrete target trajectory coordinate data to generate the reconstructed target trajectory coordinates; The three-dimensional parametric coupling module is used to map the reconstructed target trajectory coordinates to the surface topography feature matrix, extract local relative undulation data and local concave-convex curvature feature data, and calculate and generate Z-axis follow-up height command parameters and dispensing volume pulse width command parameters respectively. The underlying collaborative execution module is used to control the execution mechanism and the dispensing valve to perform three-dimensional collaborative dispensing operations based on the reconstructed target trajectory coordinates, the Z-axis follow-up height command parameters, and the dispensing volume pulse width command parameters. The trajectory reconstruction calculation module is specifically used for: calculating and generating the coordinates of the reconstructed target trajectory when: Retrieve the hand-eye calibration matrix and map the standard two-dimensional discrete target trajectory coordinate data to the pixel reference system where the interpolation direction gradient vector field is located; In the pixel reference frame, obtain the corresponding local orientation gradient vector; Based on the scaling and rotation parameters of the hand-eye calibration matrix, the local directional gradient vector is converted into a physical directional gradient vector in the mechanical motion coordinate system; The physical direction gradient vector is multiplied by a preset fluid slip compensation coefficient to generate the inverse vector space offset data defined in the mechanical motion coordinate system. The inverse vector space bias data and the standard two-dimensional discrete target trajectory coordinate data are translated and superimposed to generate the reconstructed target trajectory coordinates; The three-dimensional parametric coupling module, when extracting local relative undulation data and calculating and generating Z-axis follow-up height command parameters, is specifically used for: The reconstructed target trajectory coordinates are mapped to the pixel reference system where the surface topography feature matrix is ​​located, and the mapped pixel coordinates are used as indexes to find the corresponding local relative undulation data in the surface topography feature matrix; The local relative undulation data is multiplied by a preset height mapping coefficient and converted into a physical height offset. The physical height offset is superimposed or subtracted from the global absolute reference height parameter, the Z-axis relative installation offset, and the preset dispensing working gap to generate the Z-axis follow-up height command parameter for the current interpolation cycle. The three-dimensional parametric coupling module, when extracting local concave-convex curvature feature data and calculating and generating adhesive volume pulse width command parameters, is specifically used for: The two-dimensional Laplacian operator data corresponding to the coordinates of the reconstructed target trajectory in the surface topography feature matrix is ​​calculated using a discrete difference template, and the two-dimensional Laplacian operator data is used as the local concavity and convexity curvature feature data characterizing the local concavity and convexity polarity and steepness of the workpiece surface. The local concave-convex curvature feature data is multiplied with the preset volume compensation gain to obtain the corresponding effective level duration increment of the dispensing valve. The reference dispensing pulse width constant required for the dispensing valve to operate is added to the increment of the effective level duration to generate the dispensing volume pulse width command parameter for the current interpolation cycle.

2. The control system for a visual painting and gluing equipment for metal crafts according to claim 1, characterized in that, When the three-dimensional parametric coupling module searches for the corresponding local relative undulation data in the surface topography feature matrix, it is specifically used for: The bilinear interpolation algorithm is used to extract the integer pixel values ​​around the mapped pixel coordinates, perform weighted summation, and calculate the corresponding local relative fluctuation data. If the mapped pixel coordinates are close to the effective boundary of the surface topography feature matrix, resulting in insufficient available adjacent pixels, the nearest neighbor interpolation algorithm is switched to extract data.

3. The control system for a visual painting and adhesive application equipment for metal crafts according to claim 1, characterized in that, The height mapping coefficients are obtained through a visual offline calibration program for the step block. The acquisition process is as follows: Acquire images of stepped blocks with known height differences, and calculate the surface topography feature matrix values ​​corresponding to each stepped region; Using the known height difference as the fitting target, the height mapping coefficients used to scale the relative undulation values ​​to mechanical displacement are obtained through linear fitting or least squares fitting.

4. The control system for a visual painting and gluing equipment for metal crafts according to claim 1, characterized in that, Before calculating and generating the Z-axis follow-up height command parameter and the dispensing volume pulse width command parameter respectively, the three-dimensional parametric coupling module is also used for: The validity of the grayscale detection features at the mapped target coordinates and the two-dimensional Laplacian operator data is determined. When the gray value at the target coordinate is in the saturation range, the gray gradient difference exceeds the preset anomaly threshold, the magnitude of the local directional gradient vector exceeds the preset gradient threshold, or the two-dimensional Laplacian operator data exceeds the preset divergence threshold, the location is determined as an abnormal reflection point or an invalid detection point, and the neighborhood median filtering result, the neighborhood mean, or the data from the previous effective interpolation period are preferentially used for replacement.

5. The control system for a visual painting and adhesive application equipment for metal crafts according to claim 1, characterized in that, The underlying collaborative execution module, when controlling the actuator and dispensing valve to perform three-dimensional collaborative dispensing operations, is specifically used for: The reconstructed target trajectory coordinates are divided by the set mechanical pulse equivalent constant, and rounding is performed to convert them into the number of X-axis target pulses and the number of Y-axis target pulses that the X-axis motor and Y-axis motor can recognize. The target pulse count of the X-axis, the target pulse count of the Y-axis, the follow-up height command parameter of the Z-axis, and the dispensing volume pulse width command parameter are encapsulated into a multi-axis synchronous data packet according to the set motion control protocol; The multi-axis synchronization data packets are sent down to the underlying multi-axis servo drivers and independent dispensing controllers via industrial fieldbus to control the actuators to perform planar position interpolation and vertical micro-motion tracking, and synchronously trigger the dispensing valve to output the corresponding effective high-level duration or duty cycle electrical signal during the set trigger window period.

6. The control system for a visual painting and gluing equipment for metal crafts according to claim 5, characterized in that, Before sending the multi-axis synchronization data packet to the multi-axis servo driver and dispensing controller, the underlying collaborative execution module is also used for: Perform safety constraint judgments on the control parameters corresponding to the X-axis motor, Y-axis motor, Z-axis motor, and dispensing valve; It is determined whether the reconstructed target trajectory coordinates are within the mechanical soft limit range of the X and Y axes, whether the Z-axis follow-up height command parameter is within the safe soft limit range of the Z-axis, and whether the dispensing volume pulse width command parameter is not less than the minimum mechanical response time of the dispensing valve and not greater than the maximum allowable cycle time of the current system interpolation cycle; when any control parameter exceeds the corresponding safe range, truncation, alarm, or suspension of execution is performed.

7. A control method for a visual painting and dispensing equipment for metal crafts, applied to a control system for a visual painting and dispensing equipment for metal crafts as described in any one of claims 1-6, characterized in that, Includes the following steps: Acquire standard two-dimensional discrete target trajectory coordinate data and surface morphology feature data of the metal craft to be processed; The inverse vector space bias data corresponding to the standard two-dimensional discrete target trajectory coordinate data is calculated based on the preset interpolation direction gradient vector field, and the inverse vector space bias data is superimposed on the standard two-dimensional discrete target trajectory coordinate data to generate the reconstructed target trajectory coordinates. The reconstructed target trajectory coordinates are mapped to the surface topography feature matrix, and local relative undulation data and local concave-convex curvature feature data are extracted. The Z-axis follow-up height command parameter and the dispensing volume pulse width command parameter are calculated and generated respectively. Based on the reconstructed target trajectory coordinates, the Z-axis follow-up height command parameter, and the dispensing volume pulse width command parameter, the actuator and dispensing valve are controlled to perform a three-dimensional collaborative dispensing operation.

Images