An automatic dispensing trajectory image recognition method and system

Abstract

The present application relates to the technical field of image recognition, in particular to an automatic dispensing track image recognition method and system, the method comprising the following steps: extracting substrate pixel features, calculating edge positions, generating an affine transformation matrix, combining preset coordinates to calculate and generate a preliminary correction track, collecting environmental temperature and humidity, using a fitting function to generate a preliminary spread radius, calculating compensation instructions in combination with preset parameters, and generating adaptive compensation parameters by fusing the correction track, and adjusting motor movement and valve flow to output recognition results. In the present application, the problem of macro positioning roughness and difficulty in detecting small edge changes is effectively resolved, the contradiction between path planning static rigidity and substrate random deviation is accurately resolved, the integrated comprehensive analysis link of positioning error and process environment parameters is opened up, the device parameters are adaptively adjusted, the risk of overflow and loss of glue caused by environmental fluctuations is eliminated, and the dispensing processing precision and packaging production yield are improved.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to an automatic dispensing trajectory image recognition method and system. Background Technology

[0002] Image recognition technology mainly involves using computer hardware to extract, analyze, and understand input digital images. The core aspects of this field include image acquisition, preprocessing, feature extraction, and classification. It typically relies on camera equipment to collect visual information, followed by grayscale transformation and filtering to improve image quality, extraction of edge texture or geometric features, and finally, location and recognition of target objects in the image through template matching or connected component analysis. Among these methods, traditional automatic dispensing trajectory image recognition refers to the operation of visually locating and extracting the shape of the glue path formed by a dispensing machine applying glue to a substrate in industrial production. It typically uses an industrial camera with a ring light source to acquire grayscale images of the dispensing area and applies edge detection operators to extract the inner and outer contour pixel coordinates of the glue trajectory.

[0003] Traditional dispensing processes generally rely on fixed program control of equipment movement paths. The substrate is prone to mechanical positioning errors during the clamping process, causing the actual operation position to deviate from the expected position. Conventional vision systems are limited by pixel accuracy and cannot detect minute changes at the edges, making it impossible to build a real-time coordinate reference. The dispensing process is greatly affected by fluctuations in environmental temperature and humidity, which directly changes the physical properties and spreading pattern. Existing operation methods disconnect the comprehensive relationship between positioning deviation and environmental parameters, resulting in a static and rigid path planning characteristic. Accumulated random deviations can easily cause glue overflow or loss, thereby affecting the packaging yield. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose an automatic dispensing trajectory image recognition method and system.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an automatic dispensing trajectory image recognition method, comprising the following steps: S1: Monitor the substrate image data to extract the pixel grayscale distribution features and substrate reference geometric features of the local area of ​​the substrate, use the preset Zernike polynomial to calculate the sub-pixel position data of the substrate edge, calculate the coordinate affine mapping coefficient values ​​for the substrate edge sub-pixel position data and the preset ideal substrate coordinate data, and generate substrate affine transformation matrix data. S2: Calculate the trajectory rotation vector coordinate data based on the substrate affine transformation matrix data, the preset substrate dispensing trajectory coordinate data, and the preset substrate rotation center coordinate data, and generate preliminary corrected trajectory coordinate data; S3: Collect ambient temperature and humidity data of the dispensing area, and use the preset adhesive environment characteristic fitting function to calculate and generate the predicted adhesive spreading radius data. S4: Calculate the target adhesive dispensing volume compensation command data and the target dispensing head movement speed compensation command data using the preset ideal adhesive spreading radius data, the predicted adhesive spreading radius data, the preset basic adhesive dispensing volume data, and the preset basic dispensing head movement speed data. Then, integrate the preliminary correction trajectory coordinate data to generate dispensing trajectory adaptive compensation parameter data. S5: Adjust the spatial motion position state of the dispensing machine servo motor, the spatial motion rate state of the dispensing machine servo motor, and the glue flow output state of the dispensing valve device through the dispensing trajectory adaptive compensation parameter data to generate dispensing trajectory image recognition processing results.

[0006] The present invention is improved in that the specific steps of S1 are as follows: S111: Monitor substrate image data, extract pixel grayscale distribution features and substrate reference geometric features in local areas of the substrate, use the pixel grayscale distribution features in local areas of the substrate as the image two-dimensional function of the Zernike moment algorithm, and regard the substrate reference geometric features as the edge position of the target to be measured by the Zernike moment algorithm, and construct an image initial parameter set containing the image two-dimensional function and the edge position of the target to be measured; S112: Obtain a preset Zernike polynomial, and based on the image initial parameter set, calculate the integral projection value of the local region of the substrate according to the image two-dimensional function and the preset Zernike polynomial to obtain the substrate Zernike moment data, extract the Zernike moment phase feature and Zernike moment amplitude feature in the substrate Zernike moment data, and construct a rectangular state distribution attribute set. S113: Based on the rectangular state distribution attribute set, calculate the sub-pixel position data of the substrate edge according to the Zernike moment phase feature and the Zernike moment amplitude feature, obtain the preset ideal substrate coordinate data, calculate the coordinate affine mapping coefficient value for the substrate edge sub-pixel position data and the preset ideal substrate coordinate data, and generate substrate affine transformation matrix data.

[0007] The present invention is improved in that the specific steps of S2 are as follows: S211: For the affine transformation matrix data of the substrate, extract the current substrate origin coordinate data, the reference origin coordinate data, the current substrate coordinate axis vector data, and the reference reference coordinate axis vector data. Calculate the lateral straight line deviation value of the substrate within the horizontal axis to obtain the lateral translation deviation data of the substrate, and calculate the longitudinal straight line deviation value of the substrate within the vertical axis to obtain the longitudinal translation deviation data of the substrate. Calculate the directional angle deviation value of the substrate based on the current substrate coordinate axis vector data and the reference reference coordinate axis vector data to obtain the substrate rotation angle deviation data, and construct the substrate translation and rotation deviation parameters. S212: Obtain preset substrate dispensing trajectory coordinate data and preset substrate rotation center coordinate data; calculate trajectory relative position direction vector data based on the preset substrate dispensing trajectory coordinate data and the preset substrate rotation center coordinate data; calculate trajectory rotation vector coordinate data based on the substrate rotation angle deviation data and the trajectory relative position direction vector data within the substrate translation and rotation deviation parameters; calculate substrate spatial vector superposition sum value based on the trajectory rotation vector coordinate data, preset substrate rotation center coordinate data, substrate lateral translation deviation data, and substrate longitudinal translation deviation data; and generate preliminary corrected trajectory coordinate data.

[0008] The present invention is improved in that the process of calculating the sum of the spatial vectors of the calculation substrate to generate preliminary corrected trajectory coordinate data is specifically as follows: Extract the horizontal rotation vector coordinate values ​​and the vertical rotation vector coordinate values ​​from the vector coordinate data after the trajectory rotation; Extract the horizontal and vertical preset rotation center coordinate values ​​from the preset substrate rotation center coordinate data; The sum of the coordinate values ​​of the lateral rotation vector, the coordinate values ​​of the lateral preset rotation center, and the lateral translation deviation data of the substrate is calculated to obtain the spatial coordinate values ​​of the lateral correction trajectory. The sum of the longitudinal rotation vector coordinates, the longitudinal preset rotation center coordinates, and the longitudinal translation deviation data of the substrate is calculated to obtain the longitudinal correction trajectory spatial coordinates. A two-dimensional coordinate system mapping vector is constructed based on the horizontal and vertical corrected trajectory spatial coordinate values ​​to obtain preliminary corrected trajectory coordinate data.

[0009] The present invention is improved in that the specific steps of S3 are as follows: S311: Use temperature and humidity sensors to collect ambient temperature and humidity data of the dispensing area in the dispensing equipment, obtain a preset adhesive environment characteristic fitting function, extract temperature correlation fitting coefficient data, humidity correlation fitting coefficient data and adhesive experimental background constant data from the preset adhesive environment characteristic fitting function, use the ambient temperature data of the dispensing area as the temperature influence independent variable in the preset adhesive environment characteristic fitting function, and use the ambient humidity data of the dispensing area as the humidity influence independent variable in the preset adhesive environment characteristic fitting function, and construct a fitting parameter set containing the temperature influence independent variable and the humidity influence independent variable; S312: Call the set of fitting parameters, calculate the second power value of the independent variable itself based on the temperature-affected independent variable, calculate the weighted factor product data of the temperature independent variable based on the second power value of the independent variable itself and the temperature-related fitting coefficient data, and calculate the weighted factor product data of the humidity independent variable based on the humidity-affected independent variable and the humidity-related fitting coefficient data. S313: Calculate the sum of factors superimposed on the product data of the weighted factors of the temperature independent variable, the product data of the weighted factors of the humidity independent variable, and the data of the background constant term of the adhesive experiment, and generate the predicted adhesive spreading radius data.

[0010] The present invention is improved in that the process of calibrating and obtaining the preset adhesive environment characteristic fitting function and its included temperature correlation fitting coefficient data, humidity correlation fitting coefficient data, and adhesive experimental background constant term data is specifically as follows: Obtain the first set of adhesive spreading test samples under independent temperature change conditions and the second set of adhesive spreading test samples under independent humidity change conditions; Extract the first adhesive spreading value set corresponding to constant humidity and temperature change from the first adhesive spreading test sample set. Construct temperature correlation polynomial feature parameters based on the first adhesive spreading value set and the independent temperature change conditions. Use the temperature correlation polynomial feature parameters as the temperature correlation fitting coefficient data. Extract the second adhesive spreading value set corresponding to constant temperature and humidity change from the second adhesive spreading test sample set. Construct humidity-related linear regression feature parameters based on the second adhesive spreading value set and the independent humidity change conditions. Use the humidity-related linear regression feature parameters as the humidity-related fitting coefficient data. Extract the initial average spreading values ​​of the first adhesive spreading test sample set under the baseline experimental temperature condition and the second adhesive spreading test sample set under the baseline experimental humidity condition, and use the initial average spreading values ​​as the background constant data of the adhesive experiment; The preset adhesive environment feature fitting function is constructed based on the temperature correlation fitting coefficient data, the humidity correlation fitting coefficient data, and the adhesive experimental background constant data.

[0011] The present invention is improved in that the specific steps of S4 are as follows: S411: Obtain preset ideal adhesive spreading radius data, preset basic adhesive dispensing volume data, and preset basic dispensing head movement speed data; calculate the relative proportional quotient of adhesive spreading radius based on the preset ideal adhesive spreading radius data and the predicted adhesive spreading radius data to obtain the first scaling factor data of adhesive radius; calculate the second scaling factor data of adhesive radius based on the second scaling factor data of adhesive radius. S412: Calculate the glue volume effect multiple value based on the preset glue basic dispensing volume data and the glue radius second scaling factor data to obtain the target glue dispensing volume compensation command data; calculate the speed effect multiple value based on the preset dispensing head basic movement speed data and the glue radius first scaling factor data to obtain the target dispensing head movement speed compensation command data. S413: Perform parameter mapping and association on the preliminary corrected trajectory coordinate data, the target adhesive dispensing volume compensation command data, and the target dispensing head movement speed compensation command data to generate dispensing trajectory adaptive compensation parameter data.

[0012] The present invention is improved in that the process of generating adaptive compensation parameter data for the dispensing trajectory is specifically as follows: Extract the spatial coordinate sequence of trajectory nodes from the preliminary corrected trajectory coordinate data; Obtain the actual spatial distance value between two adjacent spatial coordinate sequences of the trajectory nodes, and calculate the smooth transition coefficient of the trajectory node by dividing the actual spatial distance value by the preset reference interpolation distance value. Multiply the target adhesive dispensing amount compensation command data by the trajectory node smoothing transition coefficient to obtain smoothed adhesive dispensing amount compensation command data. Multiply the target dispensing head motion speed compensation command data by the trajectory node smoothing transition coefficient to obtain smoothed dispensing head motion speed compensation command data. A multi-dimensional compensation data frame is constructed based on the spatial coordinate sequence of the trajectory nodes, the smoothed adhesive dispensing amount compensation instruction data, and the smoothed dispensing head movement speed compensation instruction data. The multiple multi-dimensional compensation data frames are spliced ​​together according to the time sequence to obtain the adhesive trajectory adaptive compensation parameter data.

[0013] The present invention is improved in that the specific steps of S5 are as follows: S511: Extract the spatial coordinate sequence of trajectory nodes, the smoothed dispensing head movement speed compensation command data, and the smoothed adhesive dispensing volume compensation command data from the dispensing trajectory adaptive compensation parameter data; S512: Adjust the spatial motion position state of the dispensing machine servo motor according to the spatial coordinate sequence of the trajectory nodes, adjust the spatial motion rate state of the dispensing machine servo motor according to the smoothed dispensing head motion speed compensation instruction data, and extract the state parameters after execution to construct servo motor state data. S513: Adjust the glue flow output state of the dispensing valve device according to the smoothed glue dispensing quantity compensation instruction data, extract and associate the spatial motion position state of the dispensing machine servo motor, the spatial motion speed state of the dispensing machine servo motor and the glue flow output state of the dispensing valve device from the servo motor state data, and generate the dispensing trajectory image recognition processing result.

[0014] An automatic dispensing trajectory image recognition system is provided, the automatic dispensing trajectory image recognition system being used to implement the above-mentioned automatic dispensing trajectory image recognition method, the system comprising: The visual feature analysis module monitors the substrate image data to extract the pixel grayscale distribution features and the substrate reference geometric shape features of the local area of ​​the substrate. It uses a preset Zernike polynomial to calculate the sub-pixel position data of the substrate edge and calculates the coordinate affine mapping coefficient values ​​for the sub-pixel position data of the substrate edge and the preset ideal substrate coordinate data to generate substrate affine transformation matrix data. The trajectory space synchronization module calculates the trajectory rotation vector coordinate data based on the substrate affine transformation matrix data, the preset substrate dispensing trajectory coordinate data, and the preset substrate rotation center coordinate data, and generates preliminary corrected trajectory coordinate data. The adhesive rheology prediction module collects ambient temperature and humidity data of the dispensing area, and uses a preset adhesive environment characteristic fitting function to calculate and generate the predicted adhesive spreading radius data. The multi-dimensional instruction fusion module uses preset ideal adhesive spreading radius data, predicted adhesive spreading radius data, preset basic adhesive dispensing volume data, and preset basic dispensing head movement speed data to calculate target adhesive dispensing volume compensation instruction data and target dispensing head movement speed compensation instruction data, and integrates the preliminary correction trajectory coordinate data to generate dispensing trajectory adaptive compensation parameter data. The terminal collaborative drive module adjusts the spatial motion position state of the dispensing machine servo motor, the spatial motion rate state of the dispensing machine servo motor, and the glue flow output state of the dispensing valve device through the dispensing trajectory adaptive compensation parameter data, and generates dispensing trajectory image recognition processing results.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, sub-pixel position data is calculated and an affine transformation matrix is ​​generated by extracting the local pixel grayscale distribution and reference geometric features of the substrate. The vector coordinates after rotation are specifically solved to obtain a preliminary corrected trajectory. This directly solves the problem that macroscopic positioning is too rough to detect small edge changes. It effectively breaks the contradiction between the static rigidity of path planning and the random deviation of the substrate. The ambient temperature and humidity are collected simultaneously and the fitting function is fused to calculate and predict the spreading radius. The ideal radius and basic dispensing parameters are deeply integrated to output target compensation commands. The spatial motion state of the servo motor and the flow output of the dispensing valve are adaptively adjusted. The integrated analysis link between positioning error and environmental parameters is established, eliminating the risk of overflow and loss caused by environmental fluctuations in adhesive properties. Attached Figure Description

[0016] Figure 1 This is the main flowchart for the dispensing trajectory image recognition of the present invention; Figure 2 This is a flowchart of the substrate visual feature analysis process of the present invention; Figure 3 This is a flowchart of the trajectory correction and compensation fusion process of the present invention; Figure 4 This is a flowchart of the terminal collaborative driving process of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] Please see Figures 1 to 4 This embodiment provides an automatic dispensing trajectory image recognition method and system. In practical applications, such as during the continuous operation of an electronic substrate dispensing equipment for visual positioning, trajectory correction, adhesive state compensation, and servo dispensing drive of a clamped substrate, the substrate camera continuously acquires images of the substrate surface edge and reference shape, while the dispensing area sensor simultaneously acquires temperature and humidity environmental data. The trajectory control program writes the visual recognition results, preset trajectory data, and adhesive spreading prediction results into the control parameter cache, and outputs motion and dispensing control parameters to the dispensing machine servo motor and dispensing valve equipment according to the same substrate coordinate system. The method includes the following steps: S1: Monitor the substrate image data to extract the pixel grayscale distribution features and substrate reference geometric features in the local area of ​​the substrate. Use the preset Zernike polynomial to calculate the sub-pixel position data of the substrate edge. Calculate the coordinate affine mapping coefficient values ​​for the sub-pixel position data of the substrate edge and the preset ideal substrate coordinate data to generate substrate affine transformation matrix data.

[0019] The substrate image data consists of grayscale image frames acquired by the substrate camera during the pre-dispensing positioning stage, along with associated acquisition status fields. These fields include image frame source, acquisition trigger state, camera calibration state, exposure state, and frame integrity state. The local pixel grayscale distribution characteristics of the substrate are grayscale variation fields extracted from the substrate edge neighborhood, positioning hole neighborhood, or substrate geometric reference neighborhood, carrying pixel position, grayscale level, neighborhood continuity, and noise identification. The substrate reference geometric shape characteristics are the geometric boundary fields in the image representing predefined straight edges, corners, positioning hole contours, or mounting reference contours on the substrate, used to establish a correspondence between the camera image space and the dispensing trajectory space.

[0020] The preset Zernike polynomial is an edge moment feature rule object stored in the visual positioning parameter area. This rule object records the polynomial order configuration, edge direction resolution rules, and phase amplitude association rules used for local grayscale distribution projection. The substrate edge sub-pixel position data is an edge point coordinate sequence formed after refining the substrate edge position between pixels. It includes the edge point position field, edge direction field, edge confidence state field, and anomaly identifier field in the camera plane coordinate system. The preset ideal substrate coordinate data is a set of substrate reference point coordinates saved in the dispensing process file. The coordinate affine mapping coefficient values ​​are translation, rotation, scaling, and shearing association fields formed when mapping the current substrate edge point coordinate sequence to the process reference coordinate set. The substrate affine transformation matrix data serves as the input for subsequent trajectory space synchronization processing.

[0021] Before an image frame enters step S1, its validity is determined. If the image frame is missing, the frame header is incomplete, the grayscale range falls within a camera abnormal state, the reference shape area cannot be located, or the edge continuity of a local area is insufficient, the frame is marked as an invalid image. The previous valid substrate affine transformation matrix data is inherited as temporary positioning input, and a positioning anomaly log is written. If no previous valid matrix exists during the device startup phase, the initial reference matrix data saved at the camera calibration station is called, and the dispensing trajectory is put into a pending confirmation state. The pending confirmation state is lifted after the next valid image frame is generated.

[0022] S111: Monitor substrate image data, extract pixel grayscale distribution features and substrate reference geometric features in local areas of the substrate, use the pixel grayscale distribution features in local areas of the substrate as the image two-dimensional function of the Zernike moment algorithm, and regard the substrate reference geometric features as the edge position of the target to be measured by the Zernike moment algorithm, and construct an image initial parameter set containing the image two-dimensional function and the edge position of the target to be measured.

[0023] The initial image parameter set is the input data packet for the visual positioning stage, containing fields for local grayscale distribution, target edge location, camera calibration status, substrate batch process, and preprocessing status. The local grayscale distribution field consists of the grayscale sampling location, grayscale change direction, and local noise status within the substrate edge neighborhood. The target edge location field consists of the candidate boundary region of the reference geometry, boundary direction, and connectivity status. Before entering the initial image parameter set, the grayscale distribution features undergo denoising and edge neighborhood cropping. Noise determination is based on grayscale stability rules from the camera calibration parameter area and substrate material reflectivity rules. Repeatedly acquired frames are filtered based on acquisition time order and frame integrity status, retaining only valid image frames that meet the frame integrity rules.

[0024] When local grayscale distribution features are used as a two-dimensional function of the image, the image coordinates and grayscale levels are organized into grayscale distribution objects that can participate in local integral projection, and the reference geometric features are organized into edge location objects of the target to be measured. If the reference geometry has occlusion, edge breaks, or local contamination, the preprocessing state field records the boundary identifiers of the occluded and broken areas, and only edge segments that meet the continuity rule are retained in the edge location objects of the target to be measured. The initial image parameter set is output to S112 for the Zernike moment feature extraction process to read.

[0025] S112: Obtain the preset Zernike polynomial. Based on the initial parameter set of the image, calculate the integral projection value of the local region of the substrate according to the two-dimensional function of the image and the preset Zernike polynomial to obtain the substrate Zernike moment data. Extract the Zernike moment phase feature and Zernike moment amplitude feature in the substrate Zernike moment data to construct a rectangular state distribution attribute set.

[0026] The substrate Zernike moment data is a moment feature data object formed by projecting a two-dimensional image function onto the local edge neighborhood according to a preset Zernike polynomial. It carries the directional changes, grayscale transition states, and edge morphology concentration of the local edges. The Zernike moment phase feature is used to identify the correspondence between the edge normal direction and the local grayscale change direction, while the Zernike moment amplitude feature is used to identify the concentration intensity of edge grayscale transitions within the local region. The rectangular state distribution attribute set is an intermediate data packet formed by the phase feature, amplitude feature, local edge direction state, and quality label. This data packet is stored in the visual positioning cache and associated with subsequent sub-pixel edge positioning results according to the substrate image frame identifier.

[0027] When constructing the rectangular distribution attribute set, abnormal grayscale points in the local edge neighborhood are identified by a status flag. These abnormal grayscale points include abrupt changes in grayscale caused by reflection, adhesive residue, or substrate surface contamination. Instead of being directly output as edge positions, these abnormal grayscale points are entered into the moment feature quality flag field, used to adjust the edge point confidence status in S113. If the effective edge segments within a local area are insufficient to form a continuous edge direction, the corresponding area outputs an edge candidate insufficiency status, which is then passed to S113. The coordinate affine mapping process uses the remaining effective reference points and the previous effective edge direction for constraint.

[0028] S113: Based on the rectangular state distribution attribute set, calculate the sub-pixel position data of the substrate edge according to the phase characteristics and amplitude characteristics of the Zernike moment, obtain the preset ideal substrate coordinate data, calculate the coordinate affine mapping coefficient values ​​for the sub-pixel position data of the substrate edge and the preset ideal substrate coordinate data, and generate the substrate affine transformation matrix data.

[0029] Substrate edge subpixel position data consists of edge candidate point position, edge direction, edge quality state, and reference geometry attribution fields. Phase features are used to determine the edge normal direction, and amplitude features are used to determine the confidence state of the edge in the local grayscale distribution. Together, they drive the refinement of edge candidate points from pixel-level positions to inter-pixel position fields. The preset ideal substrate coordinate data comes from the dispensing process coordinate file, carrying the substrate reference geometry in the standard clamping posture, coordinate system direction field, and reference point attribution field.

[0030] The coordinate affine mapping coefficients are generated by the correspondence between the current edge sub-pixel position data and the preset ideal substrate coordinate data. This correspondence is established through the reference geometry attribution field. Once the current edge point successfully matches the ideal reference point, a mapping field is formed to express the translation, rotation, scalability, and local deformation relationships of the current substrate relative to the process reference. If there are duplicate attributions or geometric order conflicts among the matched points, a consistency check is first performed according to the connection order of the reference geometry. Then, conflicting points are written to a low-confidence state, and the coordinate affine mapping process uses reference points whose confidence states meet the requirements to generate the substrate affine transformation matrix data. The substrate affine transformation matrix data is output to S2 and simultaneously written to the positioning traceability log. The traceability log records the image frame source, valid edge point attribution, mapping generation status, and anomaly cause fields.

[0031] S2: Calculate the vector coordinate data after trajectory rotation based on the substrate affine transformation matrix data, preset substrate dispensing trajectory coordinate data, and preset substrate rotation center coordinate data, and generate preliminary corrected trajectory coordinate data.

[0032] The preset substrate dispensing trajectory coordinate data is a sequence of spatial coordinates of trajectory nodes stored in the dispensing process file. Each trajectory node carries fields for trajectory segment affiliation, node order, process action type, and corresponding dispensing status. The preset substrate rotation center coordinate data is a rotation center field in the process coordinate system used to describe substrate posture correction. This field is determined by the substrate process reference point or clamping reference relationship and is stored in the same process coordinate system as the preset substrate dispensing trajectory coordinate data. The vector coordinate data after trajectory rotation is the intermediate coordinate field of the trajectory nodes after posture synchronization with respect to the rotation center. The preliminary corrected trajectory coordinate data is a sequence of trajectory node coordinates superimposed with substrate translational and rotational deviations, used by S4 for dispensing amount and motion speed compensation correlation.

[0033] When the substrate affine transformation matrix data is inherited from the previous valid state, S2 writes the preliminary corrected trajectory coordinate data into an inheritance state flag and passes this state to the subsequent dispensing trajectory adaptive compensation parameter data. If the inheritance state flag is not cleared, S5 reads the flag before execution and enters the confirmation or pause branch according to the equipment control rules to avoid directly driving the dispensing action without confirmed positioning data.

[0034] S211: For the affine transformation matrix data of the substrate, extract the current substrate origin coordinate data, the reference origin coordinate data, the current substrate coordinate axis vector data, and the reference reference coordinate axis vector data. Calculate the lateral linear deviation of the substrate within the horizontal axis based on the current substrate origin coordinate data and the reference reference origin coordinate data to obtain the lateral translation deviation data of the substrate. Calculate the longitudinal linear deviation of the substrate within the vertical axis to obtain the longitudinal translation deviation data of the substrate. Calculate the directional angle deviation of the substrate based on the current substrate coordinate axis vector data and the reference reference coordinate axis vector data to obtain the substrate rotation angle deviation data. Construct the substrate translation and rotation deviation parameters.

[0035] The current substrate origin coordinates and current substrate coordinate axis vectors are obtained by parsing the substrate affine transformation matrix data generated by S1. The reference origin coordinates and reference coordinate axis vectors are derived from preset ideal substrate coordinate data. The substrate lateral translation deviation data and substrate longitudinal translation deviation data are used to express the linear offset of the current substrate clamping position relative to the process reference in two axes of the planar coordinate system, respectively. The substrate rotation angle deviation data is used to express the attitude offset of the current substrate coordinate axis direction relative to the process reference coordinate axis direction.

[0036] The substrate translation and rotation deviation parameters are attitude correction data packets during the trajectory space synchronization phase, carrying fields for lateral translation deviation, longitudinal translation deviation, rotation angle deviation, deviation source status, and positioning reliability. If the current substrate coordinate axis vector data cannot be stably generated from valid edge points, the deviation source status field records insufficient axial resolution. The trajectory synchronization process calls the previous valid axial field as temporary input and marks the output trajectory as pending confirmation. If the reference reference field is missing or the process document verification fails, S211 stops outputting the substrate translation and rotation deviation parameters and writes the process reference missing status to the process document management cache. Subsequent trajectory processing does not enter the servo execution branch.

[0037] S212: Obtain preset substrate dispensing trajectory coordinate data and preset substrate rotation center coordinate data; calculate trajectory relative position direction vector data based on preset substrate dispensing trajectory coordinate data and preset substrate rotation center coordinate data; calculate trajectory rotation vector coordinate data based on substrate rotation angle deviation data within the substrate translation and rotation deviation parameters and trajectory relative position direction vector data; calculate substrate spatial vector superposition sum value based on trajectory rotation vector coordinate data, preset substrate rotation center coordinate data, substrate lateral translation deviation data and substrate longitudinal translation deviation data, and generate preliminary corrected trajectory coordinate data.

[0038] The trajectory relative position direction vector data is a field representing the direction and distance relationship between the preset trajectory nodes and the preset substrate rotation center, carrying the relative spatial position of the trajectory nodes before rotation correction. The substrate rotation angle deviation data participates in the trajectory attitude synchronization processing, ensuring that the trajectory node direction is consistent with the current substrate coordinate axis direction. After the trajectory rotation vector coordinate data is output, it is combined with the preset substrate rotation center coordinate data and substrate translation deviation data to form preliminary corrected trajectory coordinate data.

[0039] When generating preliminary corrected trajectory coordinate data, the lateral and longitudinal rotation vector coordinate values ​​are extracted from the trajectory rotation vector coordinate data, and the lateral and longitudinal preset rotation center coordinate values ​​are extracted from the preset substrate rotation center coordinate data. The lateral rotation vector coordinate values, the lateral preset rotation center coordinate values, and the substrate lateral translation deviation data are entered into the lateral corrected trajectory space coordinate field; the longitudinal rotation vector coordinate values, the longitudinal preset rotation center coordinate values, and the substrate longitudinal translation deviation data are entered into the longitudinal corrected trajectory space coordinate field. The lateral and longitudinal corrected trajectory space coordinate fields together construct a planar coordinate system mapping vector to obtain the preliminary corrected trajectory coordinate data. The preliminary corrected trajectory coordinate data retains the original sequence of trajectory nodes and the process action type, and is output to S4 as the source of the position field for the multi-dimensional compensation data frame.

[0040] S3: Collect ambient temperature and humidity data of the dispensing area, and use a preset adhesive environment characteristic fitting function to calculate and generate the predicted adhesive spreading radius data.

[0041] The ambient temperature and humidity data of the dispensing area are collected by temperature and humidity sensors installed inside the dispensing equipment. The collection trigger conditions are linked to the trajectory correction stage before dispensing, and sensor status verification, data integrity verification, and collection time alignment are completed before entering the adhesive spreading prediction stage. The preset adhesive environment characteristic fitting function is an environmental characteristic rule object formed during the adhesive process calibration stage, which carries temperature correlation fitting coefficient data, humidity correlation fitting coefficient data, and adhesive experimental background constant term data. The predicted adhesive spreading radius data is a process field that predicts the spreading range of the adhesive after the landing point under the current temperature and humidity environment, which is used by S4 to determine the adhesive quantity compensation command data and the dispensing head movement speed compensation command data.

[0042] If the temperature and humidity sensor is in a cold start state, fails the calibration, has missing data acquisition fields, or the acquisition time is not aligned with the image positioning stage, then S3 will mark the environmental data as invalid and use the most recent environmental data that meets the calibration rules as a temporary prediction input. If there is no inheritable valid environmental data, then the baseline environment field in the adhesive process file will be called, and the predicted adhesive spreading radius data will be written into the baseline environment source status. Subsequently, S4 will read this status and retain the pending confirmation flag for the compensation parameters.

[0043] S311: Use temperature and humidity sensors to collect ambient temperature and humidity data of the dispensing area within the dispensing equipment, obtain a preset adhesive environment characteristic fitting function, extract temperature correlation fitting coefficient data, humidity correlation fitting coefficient data, and adhesive experimental background constant data from the preset adhesive environment characteristic fitting function, use the ambient temperature data of the dispensing area as the temperature influence independent variable in the preset adhesive environment characteristic fitting function, and use the ambient humidity data of the dispensing area as the humidity influence independent variable in the preset adhesive environment characteristic fitting function, and construct a fitting parameter set containing the temperature influence independent variable and the humidity influence independent variable.

[0044] The fitting parameter set is the input data package for the environmental prediction stage, including fields for temperature-influenced independent variables, humidity-influenced independent variables, sensor calibration status, environmental data source status, adhesive batch process, and fitting rule version. Temperature-related fitting coefficients represent calibration correlation rules arising from temperature variations in adhesive spreading status, humidity-related fitting coefficients represent calibration correlation rules arising from humidity variations in adhesive spreading status, and adhesive experimental background constants represent the basic spreading status field formed by the adhesive under baseline calibration conditions. All of these fields are stored in the adhesive process parameter area and are jointly accessed and invoked by the adhesive material batch, dispensing valve type, and substrate surface treatment process.

[0045] When constructing the fitting parameter set, first verify whether the temperature and humidity acquisition fields come from the same dispensing area within the same device, and then verify whether the adhesive batch process field is consistent with the current dispensing task. If the adhesive batch process field does not match the fitting rule version field, the fitting parameter set is written to the rule version conflict state, and the output of valid predicted input to S312 is stopped, and the control program enters the process parameter confirmation branch. If sensor data is uploaded repeatedly, data items that meet the alignment rules are retained according to the acquisition status field and validity identifier, and duplicate records are written to the environmental data traceability log.

[0046] S312: Call the fitting parameter set, calculate the second power value of the independent variable based on the temperature-affected independent variable, calculate the weighted factor product of the temperature independent variable based on the second power value of the independent variable and the temperature-related fitting coefficient data, and calculate the weighted factor product of the humidity independent variable based on the humidity-affected independent variable and the humidity-related fitting coefficient data.

[0047] The weighted factor product data for the temperature independent variable is the temperature contribution field formed after processing the temperature-affected independent variable with temperature correlation fitting coefficient data. Similarly, the weighted factor product data for the humidity independent variable is the humidity contribution field formed after processing the humidity-affected independent variable with humidity correlation fitting coefficient data. The quadratic power processing of the temperature-affected independent variable is used to express the nonlinear response of the adhesive viscosity and spreading state to temperature changes. The processing result is not used as an independent control output but is written into the temperature contribution field and passed to S313. The humidity-affected independent variable adopts the linear calibration rule corresponding to the humidity correlation fitting coefficient data, and the processing result is written into the humidity contribution field and passed to S313.

[0048] S312 reads the source state field from the fitted parameter set. If the independent variable affecting temperature or humidity is the baseline environment source, the corresponding contribution field is written with the baseline environment source identifier. If the temperature-related or humidity-related fitting coefficient data is missing, the fitting process enters the rule missing state. S313 does not generate new predicted adhesive spreading radius data and sends a compensation pause identifier to S4. If the contribution field is successfully generated, the data, along with the adhesive experimental background constant data, enters the factor superposition and aggregation processing in S313.

[0049] S313: Calculate the sum of factors superimposed on the product data of the weighted factors of the independent variable of temperature, the product data of the weighted factors of the independent variable of humidity, and the experimental background constant data of the adhesive to generate the predicted adhesive spreading radius data.

[0050] The sum of factors is the spread prediction field, which is the result of combining the contribution fields of ambient temperature, ambient humidity, and basic adhesive spreading, all according to the same adhesive process rule. The predicted adhesive spreading radius data carries the current adhesive spreading status under the current environment, data source status, fitting rule version, and anomaly flags, and is written to the adhesive spreading prediction cache. When S4 reads the predicted adhesive spreading radius data, it simultaneously reads the data source status to determine whether subsequent compensation parameters should enter the normal output, pending confirmation output, or paused output branch.

[0051] The pre-defined adhesive environment characteristic fitting function and its included temperature-related fitting coefficient data, humidity-related fitting coefficient data, and adhesive experimental background constant term data are obtained through adhesive spreading calibration. The adhesive spreading calibration process obtains a first adhesive spreading test sample set under independent temperature variation conditions and a second adhesive spreading test sample set under independent humidity variation conditions. The first adhesive spreading test sample set provides a set of first adhesive spreading values ​​corresponding to constant humidity and temperature variation. The calibration program forms a temperature-related polynomial characteristic parameter based on the polynomial characteristic relationship between temperature change and adhesive spreading state, and uses it as the temperature-related fitting coefficient data. The second adhesive spreading test sample set provides a set of second adhesive spreading values ​​corresponding to constant temperature and humidity variation. The calibration program forms a humidity-related linear regression characteristic parameter based on the linear regression characteristic relationship between humidity change and adhesive spreading state, and uses it as the humidity-related fitting coefficient data. The initial average spreading values ​​of the first adhesive spreading test sample set under the baseline experimental temperature condition and the second adhesive spreading test sample set under the baseline experimental humidity condition form the adhesive experimental background constant term data. The calibrated temperature correlation fitting coefficient data, humidity correlation fitting coefficient data, and adhesive experimental background constant data are written into the adhesive process parameter area according to the adhesive batch, substrate surface condition, and dispensing valve equipment type, and are called by S311 when the dispensing task is started.

[0052] S4: Calculate the target adhesive dispensing volume compensation command data and the target dispensing head movement speed compensation command data using preset ideal adhesive spreading radius data, predicted adhesive spreading radius data, preset basic adhesive dispensing volume data and preset basic dispensing head movement speed data, and integrate the preliminary correction trajectory coordinate data to generate dispensing trajectory adaptive compensation parameter data.

[0053] The preset ideal adhesive spreading radius data is the target spreading state field saved in the dispensing process file. The preset basic adhesive dispensing volume data is the adhesive flow rate that the dispensing valve should output under the baseline process conditions. The preset basic dispensing head movement speed data is the speed field of the dispensing head moving along the trajectory nodes under the baseline process conditions. The target adhesive dispensing volume compensation command data is used to drive the dispensing valve equipment to correct the adhesive flow output state. The target dispensing head movement speed compensation command data is used to drive the dispensing machine servo motor to correct the trajectory movement speed state. The dispensing trajectory adaptive compensation parameter data is a control data frame sequence composed of position, speed, and dispensing volume, which is directly read by S5.

[0054] Before fusion, S4 checks whether the process task identifiers of the preliminary corrected trajectory coordinate data, predicted adhesive spreading radius data, preset ideal adhesive spreading radius data, basic adhesive dispensing volume data, and basic motion speed data are consistent. If any field originates from a pending confirmation state, the adhesive dispensing trajectory adaptive compensation parameter data is written to the pending confirmation identifier, and the process enters the control confirmation branch. If the preset process field is missing or its source is inconsistent, the generation of executable data frames is stopped, and the missing field name, task identifier, and process parameter source are written to the parameter anomaly log.

[0055] S411: Obtain preset ideal adhesive spreading radius data, preset basic adhesive dispensing volume data, and preset basic dispensing head movement speed data. Calculate the relative proportional quotient of the adhesive spreading radius based on the preset ideal adhesive spreading radius data and the predicted adhesive spreading radius data to obtain the first scaling factor data of the adhesive radius. Calculate the second scaling factor data of the adhesive radius based on the second scaling factor data of the first scaling factor data of the adhesive radius, using the second power of the scaling factor itself.

[0056] The first scaling factor data for the adhesive radius is a relative scaling field between the ideal spreading state and the predicted spreading state, used to express the direction and degree to which the adhesive spreading deviates from the process target under the current environment. The second scaling factor data for the adhesive radius is a dispensing volume correlation field further transformed from the first scaling factor, used to express the process relationship between the change in spreading radius and the effect of adhesive volume. Both scaling factors are stored as dimensionless process fields in the compensation parameter cache and are passed to S412 along with the process task identifier.

[0057] If the predicted adhesive spreading radius data contains a reference environment source identifier, the first scaling factor data and the second scaling factor data of the adhesive radius are written to the pending confirmation state. If the predicted adhesive spreading radius data contains a rule missing state, S411 stops generating scaling factors and transmits a compensation pause identifier to S412. If the preset ideal adhesive spreading radius data does not match the current adhesive batch process field, the compensation parameter cache is written to the process target conflict state, and no motion control data frames are subsequently output to the servo execution stage.

[0058] S412: Calculate the glue volume effect multiple value based on the preset glue basic dispensing volume data and the glue radius second scaling factor data to obtain the target glue dispensing volume compensation command data; calculate the speed effect multiple value based on the preset dispensing head basic movement speed data and the glue radius first scaling factor data to obtain the target dispensing head movement speed compensation command data.

[0059] The target adhesive dispensing volume compensation command data is a dispensing valve control field, including the target dispensing volume status, compensation source status, adhesive batch process identifier, and execution confirmation status. This field is generated by associating the basic dispensing volume field with the second scaling factor data of the adhesive radius, reflecting the direction of flow compensation that the dispensing valve needs to output under the current adhesive spreading prediction state. The target dispensing head movement speed compensation command data is a servo motion control field, including the target speed status, trajectory segment affiliation, compensation source status, and execution confirmation status. This field is generated by associating the basic motion speed field with the first scaling factor data of the adhesive radius, reflecting the direction of speed compensation for the dispensing head moving along the trajectory under the current adhesive spreading prediction state.

[0060] Before outputting the two target compensation commands, S412 reads the process boundary field of the dispensing valve equipment and the motion boundary field of the servo motor. The process boundary field comes from the equipment's factory configuration, dispensing valve control rules, and substrate process documents, serving as the executable boundary source for the compensation command. If the target compensation command exceeds the corresponding boundary field, the command data is written to the boundary limit state and adjusted to an executable state within the boundary according to the equipment control rules. Simultaneously, the reason for the boundary limit is written to the control traceability log. If the equipment interface returns an unexecutable state, S413 does not accept the compensation command as valid input.

[0061] S413: Perform parameter mapping and association on the preliminary correction trajectory coordinate data, the target adhesive dispensing volume compensation command data, and the target dispensing head movement speed compensation command data to generate dispensing trajectory adaptive compensation parameter data.

[0062] The spatial coordinate sequence of trajectory nodes is extracted from the preliminary corrected trajectory coordinate data. This sequence retains the node order, trajectory segment affiliation, process action type, and coordinate correction source status. The actual spatial distance between two adjacent trajectory node spatial coordinate sequences is obtained. This actual spatial distance is divided by a preset reference interpolation distance to calculate the trajectory node smoothing transition coefficient. The trajectory node smoothing transition coefficient is a process field used to smooth the movement and dispensing changes between adjacent trajectory nodes. The preset reference interpolation distance is derived from the dispensing trajectory interpolation rule table and is stored in association with the dispensing head motion control accuracy and the substrate process file.

[0063] The target adhesive dispensing volume compensation command data is multiplied by the trajectory node smoothing transition coefficient to obtain the smoothed adhesive dispensing volume compensation command data. The target dispensing head movement speed compensation command data is also multiplied by the trajectory node smoothing transition coefficient to obtain the smoothed dispensing head movement speed compensation command data. A multi-dimensional compensation data frame is constructed based on the trajectory node spatial coordinate sequence, the smoothed adhesive dispensing volume compensation command data, and the smoothed dispensing head movement speed compensation command data. The multi-dimensional compensation data frame is a control data object composed of position, speed, dispensing, status, and execution order fields under the same trajectory node. Multiple multi-dimensional compensation data frames are spliced ​​according to a time sequence to obtain the dispensing trajectory adaptive compensation parameter data.

[0064] If there are duplicate coordinates, disconnected coordinates, or discontinuous process actions between adjacent trajectory nodes, S413 will write the corresponding node into the trajectory node abnormal status. Duplicate coordinates will enter the node merging branch, disconnected coordinates will enter the trajectory interpolation confirmation branch, and discontinuous process actions will enter the action status verification branch. Trajectory segments that still cannot form a continuous data frame after abnormal branch processing will not enter the S5 execution queue, and the trajectory segment's ownership, abnormal type, and processing status will be written to the trajectory control log. Trajectory segments that can form a continuous data frame will enter the dispensing trajectory adaptive compensation parameter data and be output to S5.

[0065] S5: Adjust the spatial motion position state of the dispensing machine servo motor, the spatial motion speed state of the dispensing machine servo motor, and the glue flow output state of the dispensing valve equipment by using the dispensing trajectory adaptive compensation parameter data to generate the dispensing trajectory image recognition processing result.

[0066] Before the adaptive compensation parameter data for the dispensing trajectory enters the servo execution queue, execution confirmation is performed. This confirmation includes the trajectory node sequence status, coordinate correction source status, speed compensation status, dispensing compensation status, device interface status, and safety interlock status. The spatial motion position status is the target position control field of the dispensing machine servo motor within the substrate coordinate system; the spatial motion rate status is the speed control field of the servo motor moving along the trajectory node sequence; and the adhesive flow output status is the dispensing control field of the dispensing valve device on the corresponding trajectory node or segment. The dispensing trajectory image recognition processing result is a closed-loop result data formed by the positioning, prediction, compensation, and execution statuses, including trajectory execution status, device response status, anomaly feedback status, and log association identifiers.

[0067] If the adaptive compensation parameter data for the dispensing trajectory has a pending confirmation flag, S5 will not directly proceed with dispensing execution. Instead, it will output a manual confirmation or upper-level control confirmation status. If the device interface communication fails, the servo motor position status is not confirmed, the dispensing valve flow status is not confirmed, or the safety interlock fails, the execution queue will enter a paused state, and the current data frame position, control field source, and device return status will be written to the control traceability log. When resuming execution, the status of the most recently confirmed data frame will be read first, and then execution will continue according to the trajectory node sequence to avoid inconsistencies between the position and dispensing status caused by trajectory node jumps.

[0068] S511: Extract the spatial coordinate sequence of trajectory nodes, the smoothed dispensing head movement speed compensation command data, and the smoothed adhesive dispensing volume compensation command data from the dispensing trajectory adaptive compensation parameter data.

[0069] S511 reads the dispensing trajectory adaptive compensation parameter data according to the data frame execution order, writes the trajectory node spatial coordinate sequence into the position control cache, writes the smoothed dispensing head movement speed compensation instruction data into the speed control cache, and writes the smoothed adhesive dispensing volume compensation instruction data into the dispensing control cache. Each cache entry carries the same data frame identifier, trajectory segment affiliation, and execution confirmation status, ensuring that the position, speed, and dispensing fields are synchronously accessed under the same trajectory node.

[0070] If, during the reading process, a data frame identifier is missing, the position and speed fields are inconsistent, the glue dispensing field does not match the trajectory action type, or the execution order field is broken, the corresponding data frame enters a frame consistency anomaly state. After the frame consistency anomaly state is written to the control buffer, S512 and S513 will not read the frame as a valid control input and will report the anomaly to the trajectory control log of S413. Data frames that meet the consistency rules enter the execution input queues of S512 and S513.

[0071] S512: Adjust the spatial motion position state of the dispensing machine servo motor according to the spatial coordinate sequence of the trajectory nodes, adjust the spatial motion rate state of the dispensing machine servo motor according to the smoothed dispensing head motion speed compensation instruction data, and extract the state parameters after execution to construct servo motor state data.

[0072] After the spatial coordinate sequence of trajectory nodes enters the servo motor position control interface, the servo drive control generates position execution states according to the node sequence. These positions include target position reception state, motion path entry state, node arrival state, and position feedback state. After the smoothed dispensing head motion speed compensation command data enters the servo motor speed control interface, the speed execution states include speed command reception state, speed limit state, trajectory segment speed hold state, and speed feedback state. The servo motor status data is an execution feedback data object composed of position execution state, speed execution state, data frame identifier, and device feedback identifier, used by the S513 to associate the dispensing valve output status.

[0073] If the servo motor returns to an unconfirmed position or speed state, the servo motor status data is written to an unconfirmed execution flag, and the dispensing valve dispensing action of the corresponding data frame is paused. If the servo motor returns to a boundary limit state, the boundary limit reason is written to the speed control buffer and position control buffer, and the adjusted executable state is fed back to the dispensing trajectory image recognition processing result. If the feedback state after the servo motor completes execution is inconsistent with the data frame flag, the control process enters the frame feedback mismatch branch and stops associating the dispensing state.

[0074] S513: Adjust the glue flow output status of the dispensing valve device according to the smoothed glue dispensing quantity compensation instruction data, extract and associate the spatial motion position status of the dispensing machine servo motor, the spatial motion speed status of the dispensing machine servo motor and the glue flow output status of the dispensing valve device from the servo motor status data, and generate the dispensing trajectory image recognition processing result.

[0075] After the smoothed adhesive dispensing volume compensation command data enters the dispensing valve control interface, the dispensing valve device confirms the dispensing timing based on the data frame identifier, trajectory segment assignment, and servo motor status data, and outputs the adhesive flow status. The adhesive flow output status includes dispensing command reception status, valve opening status, flow holding status, flow feedback status, and abnormal feedback status. Only when both the position and speed status in the servo motor status data are in an execution confirmation state is the adhesive flow output status of the dispensing valve device associated with the corresponding data frame.

[0076] The dispensing trajectory image recognition processing result consists of substrate affine transformation matrix data, preliminary corrected trajectory coordinate data, predicted adhesive spreading radius data, dispensing trajectory adaptive compensation parameter data, servo motor status data, and adhesive flow output status. This result is written to the task result cache and traceability log. The traceability log records the image frame source, environmental data source, trajectory node affiliation, compensation status, equipment feedback status, and anomaly handling status. If the dispensing valve returns an unconfirmed flow status, the result data is written to a dispensing anomaly flag and fed back to the control confirmation branch. If the position, speed, and flow rates are all correlated, the result data is written to the trajectory execution completion status and serves as the source of the previous valid control status for the next substrate positioning cycle.

[0077] The automatic dispensing trajectory image recognition system in this embodiment is used to implement the aforementioned automatic dispensing trajectory image recognition method. The system includes a visual feature analysis module, a trajectory space synchronization module, an adhesive rheology prediction module, a multi-dimensional instruction fusion module, and a terminal collaborative driving module. The visual feature analysis module receives substrate image data output from the substrate camera, performs grayscale distribution extraction, reference geometry recognition, Zernike moment feature processing, and substrate affine transformation matrix data generation in steps S1 to S113, and outputs the matrix data and positioning status to the trajectory space synchronization module. The trajectory space synchronization module reads the substrate affine transformation matrix data output from the visual feature analysis module, as well as the preset trajectory data and rotation center data in the process file, and performs trajectory attitude synchronization and preliminary correction trajectory coordinate data generation in steps S2 to S212. The adhesive rheology prediction module receives environmental fields collected by the temperature and humidity sensor and fitting rule fields in the adhesive process parameter area, and performs environmental data verification, adhesive spreading prediction, and prediction status output in steps S3 to S313. The multi-dimensional instruction fusion module reads the preliminary correction trajectory coordinate data, the predicted adhesive spreading radius data, and the basic process control fields, and carries out the dispensing volume compensation, speed compensation, trajectory node smoothing, and multi-dimensional compensation data frame stitching from S4 to S413. The terminal collaborative drive module connects to the control interface of the dispensing machine servo motor and dispensing valve equipment, and carries out the position status adjustment, speed status adjustment, adhesive flow status adjustment, execution feedback association, and dispensing trajectory image recognition processing result generation from S5 to S513. All modules transmit data through the same process task identifier and data frame identifier. Abnormal states are written to the traceability log by the generating module, and the receiving module enters the corresponding confirmation, pause, or inheritance branch when reading.

[0078] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An automatic dispensing trajectory image recognition method, characterized in that, Includes the following steps: S1: Monitor the substrate image data to extract the pixel grayscale distribution features and substrate reference geometric features of the local area of ​​the substrate, use the preset Zernike polynomial to calculate the sub-pixel position data of the substrate edge, calculate the coordinate affine mapping coefficient values ​​for the substrate edge sub-pixel position data and the preset ideal substrate coordinate data, and generate substrate affine transformation matrix data. S2: Calculate the trajectory rotation vector coordinate data based on the substrate affine transformation matrix data, the preset substrate dispensing trajectory coordinate data, and the preset substrate rotation center coordinate data, and generate preliminary corrected trajectory coordinate data; S3: Collect ambient temperature and humidity data of the dispensing area, and use the preset adhesive environment characteristic fitting function to calculate and generate the predicted adhesive spreading radius data. S4: Calculate the target adhesive dispensing volume compensation command data and the target dispensing head movement speed compensation command data using the preset ideal adhesive spreading radius data, the predicted adhesive spreading radius data, the preset basic adhesive dispensing volume data, and the preset basic dispensing head movement speed data. Then, integrate the preliminary correction trajectory coordinate data to generate dispensing trajectory adaptive compensation parameter data. S5: Adjust the spatial motion position state of the dispensing machine servo motor, the spatial motion rate state of the dispensing machine servo motor, and the glue flow output state of the dispensing valve device through the dispensing trajectory adaptive compensation parameter data to generate dispensing trajectory image recognition processing results.

2. The automatic dispensing trajectory image recognition method according to claim 1, characterized in that, The specific steps of S1 are as follows: S111: Monitor substrate image data, extract pixel grayscale distribution features and substrate reference geometric features in local areas of the substrate, use the pixel grayscale distribution features in local areas of the substrate as the image two-dimensional function of the Zernike moment algorithm, and regard the substrate reference geometric features as the edge position of the target to be measured by the Zernike moment algorithm, and construct an image initial parameter set containing the image two-dimensional function and the edge position of the target to be measured; S112: Obtain a preset Zernike polynomial, and based on the image initial parameter set, calculate the integral projection value of the local region of the substrate according to the image two-dimensional function and the preset Zernike polynomial to obtain the substrate Zernike moment data, extract the Zernike moment phase feature and Zernike moment amplitude feature in the substrate Zernike moment data, and construct a rectangular state distribution attribute set. S113: Based on the rectangular state distribution attribute set, calculate the sub-pixel position data of the substrate edge according to the Zernike moment phase feature and the Zernike moment amplitude feature, obtain the preset ideal substrate coordinate data, calculate the coordinate affine mapping coefficient value for the substrate edge sub-pixel position data and the preset ideal substrate coordinate data, and generate substrate affine transformation matrix data.

3. The automatic dispensing trajectory image recognition method according to claim 1, characterized in that, The specific steps of S2 are as follows: S211: For the affine transformation matrix data of the substrate, extract the current substrate origin coordinate data, the reference origin coordinate data, the current substrate coordinate axis vector data, and the reference reference coordinate axis vector data. Calculate the lateral straight line deviation value of the substrate within the horizontal axis to obtain the lateral translation deviation data of the substrate, and calculate the longitudinal straight line deviation value of the substrate within the vertical axis to obtain the longitudinal translation deviation data of the substrate. Calculate the directional angle deviation value of the substrate based on the current substrate coordinate axis vector data and the reference reference coordinate axis vector data to obtain the substrate rotation angle deviation data, and construct the substrate translation and rotation deviation parameters. S212: Obtain preset substrate dispensing trajectory coordinate data and preset substrate rotation center coordinate data; calculate trajectory relative position direction vector data based on the preset substrate dispensing trajectory coordinate data and the preset substrate rotation center coordinate data; calculate trajectory rotation vector coordinate data based on the substrate rotation angle deviation data and the trajectory relative position direction vector data within the substrate translation and rotation deviation parameters; calculate substrate spatial vector superposition sum value based on the trajectory rotation vector coordinate data, preset substrate rotation center coordinate data, substrate lateral translation deviation data, and substrate longitudinal translation deviation data; and generate preliminary corrected trajectory coordinate data.

4. The automatic dispensing trajectory image recognition method according to claim 3, characterized in that, The process of calculating the sum of the spatial vectors of the computational substrate to generate preliminary corrected trajectory coordinate data is as follows: Extract the horizontal rotation vector coordinate values ​​and the vertical rotation vector coordinate values ​​from the vector coordinate data after the trajectory rotation; Extract the horizontal and vertical preset rotation center coordinate values ​​from the preset substrate rotation center coordinate data; The sum of the coordinate values ​​of the lateral rotation vector, the coordinate values ​​of the lateral preset rotation center, and the lateral translation deviation data of the substrate is calculated to obtain the spatial coordinate values ​​of the lateral correction trajectory. The sum of the longitudinal rotation vector coordinates, the longitudinal preset rotation center coordinates, and the longitudinal translation deviation data of the substrate is calculated to obtain the longitudinal correction trajectory spatial coordinates. Based on the horizontal and vertical corrected trajectory spatial coordinate values, a two-dimensional coordinate system mapping vector is constructed to obtain preliminary corrected trajectory coordinate data.

5. The automatic dispensing trajectory image recognition method according to claim 1, characterized in that, The specific steps of S3 are as follows: S311: Use temperature and humidity sensors to collect ambient temperature and humidity data of the dispensing area in the dispensing equipment, obtain a preset adhesive environment characteristic fitting function, extract temperature correlation fitting coefficient data, humidity correlation fitting coefficient data and adhesive experimental background constant data from the preset adhesive environment characteristic fitting function, use the ambient temperature data of the dispensing area as the temperature influence independent variable in the preset adhesive environment characteristic fitting function, and use the ambient humidity data of the dispensing area as the humidity influence independent variable in the preset adhesive environment characteristic fitting function, and construct a fitting parameter set containing the temperature influence independent variable and the humidity influence independent variable; S312: Call the set of fitting parameters, calculate the second power value of the independent variable itself based on the temperature-affected independent variable, calculate the weighted factor product data of the temperature independent variable based on the second power value of the independent variable itself and the temperature-related fitting coefficient data, and calculate the weighted factor product data of the humidity independent variable based on the humidity-affected independent variable and the humidity-related fitting coefficient data. S313: Calculate the sum of factors superimposed on the product data of the weighted factors of the temperature independent variable, the product data of the weighted factors of the humidity independent variable, and the data of the background constant term of the adhesive experiment, and generate the predicted adhesive spreading radius data.

6. The automatic dispensing trajectory image recognition method according to claim 5, characterized in that, The specific process for obtaining the calibration of the preset adhesive environment characteristic fitting function and its included temperature correlation fitting coefficient data, humidity correlation fitting coefficient data, and adhesive experimental background constant data is as follows: Obtain the first set of adhesive spreading test samples under independent temperature change conditions and the second set of adhesive spreading test samples under independent humidity change conditions; Extract the first adhesive spreading value set corresponding to constant humidity and temperature change from the first adhesive spreading test sample set. Construct temperature correlation polynomial feature parameters based on the first adhesive spreading value set and the independent temperature change conditions. Use the temperature correlation polynomial feature parameters as the temperature correlation fitting coefficient data. Extract the second adhesive spreading value set corresponding to constant temperature and humidity change from the second adhesive spreading test sample set. Construct humidity-related linear regression feature parameters based on the second adhesive spreading value set and the independent humidity change conditions. Use the humidity-related linear regression feature parameters as the humidity-related fitting coefficient data. Extract the initial average spreading values ​​of the first adhesive spreading test sample set under the baseline experimental temperature condition and the second adhesive spreading test sample set under the baseline experimental humidity condition, and use the initial average spreading values ​​as the background constant data of the adhesive experiment; The preset adhesive environment feature fitting function is constructed based on the temperature correlation fitting coefficient data, the humidity correlation fitting coefficient data, and the adhesive experimental background constant data.

7. The automatic dispensing trajectory image recognition method according to claim 1, characterized in that, The specific steps of S4 are as follows: S411: Obtain preset ideal adhesive spreading radius data, preset basic adhesive dispensing volume data, and preset basic dispensing head movement speed data; calculate the relative proportional quotient of adhesive spreading radius based on the preset ideal adhesive spreading radius data and the predicted adhesive spreading radius data to obtain the first scaling factor data of adhesive radius; calculate the second scaling factor data of adhesive radius based on the second scaling factor data of adhesive radius. S412: Calculate the glue volume effect multiple value based on the preset glue basic dispensing volume data and the glue radius second scaling factor data to obtain the target glue dispensing volume compensation command data; calculate the speed effect multiple value based on the preset dispensing head basic movement speed data and the glue radius first scaling factor data to obtain the target dispensing head movement speed compensation command data. S413: Perform parameter mapping and association on the preliminary corrected trajectory coordinate data, the target adhesive dispensing volume compensation command data, and the target dispensing head movement speed compensation command data to generate dispensing trajectory adaptive compensation parameter data.

8. The automatic dispensing trajectory image recognition method according to claim 7, characterized in that, The process of generating adaptive compensation parameter data for the dispensing trajectory is as follows: Extract the spatial coordinate sequence of trajectory nodes from the preliminary corrected trajectory coordinate data; Obtain the actual spatial distance value between two adjacent spatial coordinate sequences of the trajectory nodes, and calculate the smooth transition coefficient of the trajectory node by dividing the actual spatial distance value by the preset reference interpolation distance value. Multiply the target adhesive dispensing amount compensation command data by the trajectory node smoothing transition coefficient to obtain smoothed adhesive dispensing amount compensation command data. Multiply the target dispensing head motion speed compensation command data by the trajectory node smoothing transition coefficient to obtain smoothed dispensing head motion speed compensation command data. A multi-dimensional compensation data frame is constructed based on the spatial coordinate sequence of the trajectory nodes, the smoothed adhesive dispensing amount compensation instruction data, and the smoothed dispensing head movement speed compensation instruction data. The multiple multi-dimensional compensation data frames are spliced ​​together according to the time sequence to obtain the adhesive trajectory adaptive compensation parameter data.

9. The automatic dispensing trajectory image recognition method according to claim 1, characterized in that, The specific steps of S5 are as follows: S511: Extract the spatial coordinate sequence of trajectory nodes, the smoothed dispensing head movement speed compensation command data, and the smoothed adhesive dispensing volume compensation command data from the dispensing trajectory adaptive compensation parameter data; S512: Adjust the spatial motion position state of the dispensing machine servo motor according to the spatial coordinate sequence of the trajectory nodes, adjust the spatial motion rate state of the dispensing machine servo motor according to the smoothed dispensing head motion speed compensation instruction data, and extract the state parameters after execution to construct servo motor state data. S513: Adjust the glue flow output state of the dispensing valve device according to the smoothed glue dispensing quantity compensation instruction data, extract and associate the spatial motion position state of the dispensing machine servo motor, the spatial motion speed state of the dispensing machine servo motor and the glue flow output state of the dispensing valve device from the servo motor state data, and generate the dispensing trajectory image recognition processing result.

10. An automatic dispensing trajectory image recognition system, characterized in that, The system is used to implement the automatic dispensing trajectory image recognition method as described in any one of claims 1-9, the system comprising: The visual feature analysis module monitors the substrate image data to extract the pixel grayscale distribution features and the substrate reference geometric shape features of the local area of ​​the substrate. It uses a preset Zernike polynomial to calculate the sub-pixel position data of the substrate edge and calculates the coordinate affine mapping coefficient values ​​for the sub-pixel position data of the substrate edge and the preset ideal substrate coordinate data to generate substrate affine transformation matrix data. The trajectory space synchronization module calculates the trajectory rotation vector coordinate data based on the substrate affine transformation matrix data, the preset substrate dispensing trajectory coordinate data, and the preset substrate rotation center coordinate data, and generates preliminary corrected trajectory coordinate data. The adhesive rheology prediction module collects ambient temperature and humidity data of the dispensing area, and uses a preset adhesive environment characteristic fitting function to calculate and generate the predicted adhesive spreading radius data. The multi-dimensional instruction fusion module uses preset ideal adhesive spreading radius data, predicted adhesive spreading radius data, preset basic adhesive dispensing volume data, and preset basic dispensing head movement speed data to calculate target adhesive dispensing volume compensation instruction data and target dispensing head movement speed compensation instruction data, and integrates the preliminary correction trajectory coordinate data to generate dispensing trajectory adaptive compensation parameter data. The terminal collaborative drive module adjusts the spatial motion position state of the dispensing machine servo motor, the spatial motion rate state of the dispensing machine servo motor, and the glue flow output state of the dispensing valve device through the dispensing trajectory adaptive compensation parameter data, and generates dispensing trajectory image recognition processing results.

Images