Foreign matter detection visual identification system for display panel production

Abstract

The invention discloses a foreign matter detection visual identification system for display panel production, and relates to the technical field of display production, and the system comprises a monitoring area division module which divides a display panel into a plurality of monitoring sub-areas based on a display panel structure diagram and a process flow diagram, each monitoring sub-area is endowed with a unique area number, and the area number of each monitoring sub-area is stored; the label comprises a region coordinate range, a structure label and a process stage label; and the identification detection module is used for carrying out image detection on each monitoring sub-region, acquiring a local image feature set, carrying out enhancement processing on the local image feature set, and outputting a group of standardized structure enhanced image identification data. According to the invention, a multi-dimensional process parameter sequence modeling mechanism is introduced into the mapping module, and depth time sequence prediction models such as a long-short-term memory network and the like are integrated, so that the possibility of occurrence of foreign matter defects in the future can be predicted based on parameter variation trends such as temperature, pressure and exposure power of each monitoring sub-region in different process stages.

Description

Technical Field

[0001] The present invention relates to the technical field of display production, and in particular to a foreign matter detection and visual recognition system for display panel production. Background Art

[0002] With the large-scale application of new panels such as high-resolution displays and flexible OLEDs, the requirements for cleanliness and process stability in the panel production process are constantly increasing. Because display panels involve complex processes such as material lamination, temperature control, and stress release during multiple processing steps such as lamination, exposure, and curing, they are prone to introducing foreign matter defects such as bubbles, particles, filaments, and film warping. These foreign matter not only affects product yield but may also cause quality issues such as bright spots, dark shadows, and color spots during subsequent use. Therefore, how to accurately identify these foreign matter and track their causes has become a key issue that needs to be addressed in the panel manufacturing industry.

[0003] A Chinese patent (publication number: CN1779473A) discloses a method and apparatus for inspecting flat-panel displays using a visual model. The patent's steps include: capturing an image of the panel to be tested; adjusting the captured image to generate an image to be tested and a reference background image; subjecting the image to be tested and the reference image to a recognition process that includes pre-processing, estimating image information, and integrating image information; finally, combining the test information to generate a test image and a test value; and using the test value and the test image to assess the image quality of the flat-panel display to determine the quality of the panel.

[0004] Current technology often uses methods based on visual image recognition to detect defects in display panels. These methods, including but not limited to overhead industrial cameras, multi-angle light source systems, image enhancement algorithms, and artificial intelligence recognition models, can identify significant foreign matter to a certain extent. However, their effectiveness remains limited when it comes to identifying some optically weak foreign matter (such as translucent particles and films with raised edges) or defects buried within multi-layer structures. Furthermore, while some systems can record image anomalies, they struggle to effectively link them to specific process steps and equipment numbers, hindering subsequent fault analysis and precise rectification. Therefore, the present invention proposes a foreign matter detection visual recognition system for display panel production. Summary of the Invention

[0005] The object of the present invention is to provide a foreign body detection and visual recognition system for display panel production to solve the problems mentioned in the above background technology.

[0006] The present invention can be implemented by the following technical solutions: a foreign body detection and visual recognition system for display panel production, comprising a monitoring area division module, an identification and detection module, a mapping module, and an analysis and processing module;

[0007] The monitoring area division module is based on the display panel structure diagram and the process flow diagram, and divides the display panel into multiple monitoring sub-areas with physical boundaries and functional distinctions according to the hierarchical structure of the display panel;

[0008] Each monitoring sub-area is assigned a unique area number;

[0009] The area numbering is synchronized with the production line rolling code, and the image frame and physical panel position consistency are achieved through the feature point alignment algorithm;

[0010] When speed fluctuations, camera acquisition delays, and other issues occur during production line operation, causing the corresponding positions of the monitoring sub-areas in the image to shift, the system calculates the drift value based on the structural points identified in the image frame and dynamically corrects the area number to achieve self-correction of the number.

[0011] The recognition and detection module uses the microstructural differences of the display panel as input, and dynamically configures the image acquisition method and imaging parameters by sensing the physical structural properties, process state characteristics, and potential foreign matter types of the area to be tested, so that the potential foreign matter in the monitored sub-area is optimally responded to in the image.

[0012] The recognition and detection module detects the monitoring sub-area through multi-view image fusion, multi-angle illumination scanning and disparity map construction, obtains a local image feature set, performs enhancement processing, and outputs a set of standardized structure-enhanced image recognition data. The local image feature set is the identifiable visual response features such as brightness, contrast, edge, texture, etc. of the corresponding area in the image formed by the foreign matter under specific image acquisition conditions on the display panel surface or structural layer;

[0013] The structural enhancement image recognition data includes: the foreign body enhanced image of the corresponding monitoring sub-area, the spatial location of the foreign body within the area, the foreign body morphology classification result, the image enhancement method label, the recognition confidence score and the area number information, so as to ensure that foreign bodies with different attachment methods, different structural hierarchical positions and different orientation characteristics can be recognized with high confidence under the unified enhancement recognition process;

[0014] The structurally enhanced image recognition data not only improves the visibility of foreign object images, but also retains the positioning information of foreign objects in the spatial structure of the display panel, which is used to support subsequent defect classification, spatial coordinate calculation and process traceability analysis.

[0015] The mapping module is used to identify the association and binding between the structure-enhanced image recognition data output by the detection module and the corresponding processing technology, production equipment and historical parameters, and to build a source traceability path for the foreign body recognition results;

[0016] The mapping module receives the recognition result including the area number information, and based on the mapping relationship between the area number and the production process code, queries and extracts the processing steps, processing equipment numbers, equipment operation time periods and corresponding key process parameters corresponding to the monitoring sub-area;

[0017] Key process parameters include, but are not limited to, coating thickness (μm), exposure energy (mJ / cm²), lamination pressure (kgf), UV irradiation intensity (W / cm²), lamination temperature (°C), and ambient temperature and humidity during processing (%RH, °C). These process parameters are provided by the sensor network or MES system during the production process and are matched and bound to the monitoring sub-area number using timestamps and panel numbers.

[0018] The mapping module further establishes a multi-field mapping table containing "monitoring sub-area number - processing equipment number - process parameter set - timestamp" to accurately map each defect data point in the foreign body identification results to its potential processing source, providing input basis for equipment responsibility confirmation, process parameter traceability, and subsequent analysis and processing;

[0019] The analysis and processing module is used to fuse the structure-enhanced image recognition data output by the recognition and detection module with the processing traceability data generated by the mapping module to perform defect risk assessment, quality anomaly trend analysis and cross-regional defect association judgment.

[0020] A further technical improvement of the present invention is that the method for dynamically correcting the area number by the system includes the following steps:

[0021] S1. Determine whether there is a risk of number misalignment:

[0022] The system calculates the time difference between the acquisition time of the image frame and the time when the region number is bound;

[0023] If the theoretical production line movement distance corresponding to the time difference exceeds the preset area length threshold, it is preliminarily determined that the current image may have a number misalignment and the correction process is initiated;

[0024] S2. Extract structural feature points in the image:

[0025] Perform edge detection and feature point extraction on the current image frame to identify key structural points in the image, such as display border lines, positioning corners, fitting frame boundaries, and other parts with strong structural stability and identifiability, as a reference for determining the actual position of the image;

[0026] S3. Match with standard structure template:

[0027] The system extracts standard points from the standard structure template corresponding to the corresponding area number, compares them with the key structure points extracted from the image, and uses boundary contour matching, feature point pairing or image registration methods (such as affine transformation fitting) to calculate the offset between the key structure points and the standard point positions to obtain the image drift distance and direction;

[0028] S4. Determine whether the number needs to be corrected:

[0029] The system sets an offset tolerance threshold. If the actual position of the image structure point deviates from the template position by more than the tolerance threshold, the current image number is considered inconsistent with the actual area. Based on the direction and distance of the offset, the system calculates how many numbers should be adjusted forward or backward.

[0030] S5. Execution number correction:

[0031] Based on the offset judgment result, the system replaces the original number of the current image frame with the corrected area number, and records the correction information in the number correction table, including the original number, new number, drift distance, correction time, etc.

[0032] S6. Closed-loop feedback and subsequent processing:

[0033] The correction information is fed back to the monitoring area division module. If the number offset occurs in multiple consecutive image frames, the system will output an alarm message to indicate that there is a synchronization anomaly;

[0034] At the same time, the corrected number is used for subsequent attribution binding of identification data to ensure that the identification results are correctly classified into their actual structural areas.

[0035] A further technical improvement of the present invention is that: the recognition and detection module includes a lateral viewing angle detection subunit, an adjustable optical axis deflection subunit and a parallax enhancement recognition subunit;

[0036] The above subunits are deployed simultaneously in the system structure, and the system can selectively activate one or more subunits to perform image acquisition and enhancement processing based on the predicted results of the monitoring sub-area structure, process status or foreign matter type.

[0037] A further technical improvement of the present invention is that the side viewing angle detection subunit is used to identify non-frontal visible defects in the edge area or structural parts of the display panel due to poor bonding, loose structure or micro-foreign matter adhesion, including residual bonding colloid, edge film material warping, structural component bulging, etc.

[0038] Specifically, the lateral viewing angle detection subunit sets up multiple non-vertically mounted industrial cameras in a specific monitoring sub-area of ​​the display panel, and a fixed tilt angle is formed between the optical axis of each camera and the normal direction of the panel, and the tilt angle range is 30° to 60°;

[0039] The layout locations include but are not limited to:

[0040] The four border areas of the panel (used to detect the edge sealing structure);

[0041] Polarizer lamination edge (for detecting laminated foreign matter);

[0042] FPC bonding area (used to detect foreign matter at the side outlet);

[0043] At the same time, a main downward-looking camera is placed directly above the panel, forming a joint acquisition array with the above-mentioned side cameras;

[0044] The workflow of the lateral view detection subunit includes:

[0045] Image acquisition: When the target area enters the inspection station, the system synchronously triggers the top-view camera and all side-view cameras to acquire images, obtaining top-view images and side-view images respectively, ensuring that the image timestamps are consistent;

[0046] Spatial registration: Perform affine transformation and angle correction on the top view image and each side view image. Specifically, this includes: extracting the image edge point set and performing preliminary registration with the boundary features in the structural template; solving the affine matrix using the edge corner points and performing coordinate remapping; and uniformly transforming the top view image and each side view image into a regional image in a standard reference coordinate system.

[0047] Image fusion and contrast enhancement: Perform pixel-weighted fusion on the remapped images from each perspective to obtain a fused image and enhance edge structure signals;

[0048] Detect the grayscale variability of local areas in the fused image to generate a preliminary heat map of suspected foreign bodies;

[0049] Edge defect feature extraction: In the fused image, select the preset corner area ROI and perform: Canny edge extraction; local texture directional variation detection based on Gabor filter; connection analysis and morphological filtering to extract abnormal contour areas;

[0050] Defect determination: Based on contour integrity, texture interference index, and structural symmetry indicators, a classification model is constructed to classify and determine the following types of defects that may exist in corner areas.

[0051] A further technical improvement of the present invention is that the adjustable optical axis deflection sub-unit is used to improve the visibility of slightly tilted, weakly adhered, or low-reflectivity foreign objects on the display panel surface in the image. It is particularly suitable for detecting lightly adhered particulate impurities, tilted adhesive residue, transparent foreign objects, and other defective objects that are difficult to image under conventional vertical incident light, including but not limited to the following hardware structure:

[0052] Controllable deflection component: preferably using a MEMS micromirror array, a dual-axis servo pan / tilt, or an electric mirror mechanism to precisely control the incident direction of light, with a deflection angle range of ±30°;

[0053] Directional light source: emits a focusable and controllable directional lighting beam, coupled with a deflection component to ensure local lighting at any angle;

[0054] Optical axis control driver: with high-precision angle control capability, the angle adjustment step accuracy is preferably 5°;

[0055] Incident angle database: stores the recommended incident angle range and preferred angle value associated with each monitoring sub-area number, which is obtained based on historical training data, structural morphology and material reflection characteristics;

[0056] Its workflow includes:

[0057] Incident angle configuration acquisition: When a certain area number i enters the current detection station, the system reads the incident angle setting value set corresponding to the area from the incident angle database ;

[0058] Angle scanning acquisition: Control the controllable deflection component to set the incident angle one by one to the collection Various angles in ;

[0059] Each angle set , that is, synchronously trigger the industrial camera to collect images , until all angle images are collected and the image sequence is obtained ={ };

[0060] Image enhancement processing: image sequence Perform local contrast enhancement processing (such as CLAHE, gradient boosting) on ​​each frame of the image and evaluate the response intensity of the foreign body target area ;

[0061] If there is a strongest response image , then the frame image is selected as the final enhanced image;

[0062] If the difference in the response intensity of n frames is less than the preset difference threshold, the image stacking and fusion processing is performed to generate the final enhanced image ;

[0063] Foreign body recognition execution: the image or Input the preset defect recognition network to classify and locate weak reflectors, low-contrast particles, and semi-transparent areas, and output the defect type, location, area, and reflection intensity labels.

[0064] A further technical improvement of the present invention is that the parallax enhancement recognition subunit is used to identify slender foreign objects (such as filaments, hair, and thin fibers) on the display panel surface that have strong directionality and are axially parallel to the main line of sight. This solves the detection blind spot problem of such foreign objects in vertical top-down imaging, where the reflection is weak, the shape is blurred, and the boundaries are unclear. Specifically:

[0065] The parallax enhancement recognition subunit consists of two industrial cameras positioned on either side of the target monitoring subarea of ​​the display panel. Both cameras are tilted ±15° to the left and right relative to the panel normal, ensuring their line of sight is horizontal. Both cameras are equipped with synchronized triggering devices to ensure consistent image acquisition timing and eliminate motion interference. The geometric relationship between the cameras is determined through a post-installation calibration process, including their respective intrinsic parameter matrices and distortion coefficients (for distortion correction), as well as their rotational and translational relationships (for geometric calibration).

[0066] Its workflow includes synchronous image acquisition, disparity map construction, edge difference enhancement, and directional contour fitting and feature extraction;

[0067] Synchronous image acquisition:

[0068] When the detection station reaches the target area, the two cameras belonging to the parallax enhancement recognition subunit respectively capture the image of the current monitoring sub-area, which is recorded as the left view. and right view , and the parallax enhancement recognition subunit for the left view and right view Perform image distortion correction and geometric calibration to obtain the left view and right view Paired images of

[0069] Geometric calibration refers to the system's calibration of the left view based on the installation angle and relative position (i.e., external parameters) between the cameras. and right view Perform geometric calibration and registration transformations to align the two images in spatial structure, ensuring that the pixels of the targets in the images are completely aligned in the vertical direction when calculating the disparity, and only retaining the disparity difference in the horizontal direction, thus forming a standard image pairing relationship;

[0070] Disparity map construction:

[0071] Left view and right view Execute stereo matching algorithms based on block matching or semi-global matching;

[0072] Output a disparity image D, where the grayscale value of each pixel in the disparity image D represents the horizontal offset of that pixel between the two images. The disparity image D can highlight structural features with slight height variations or texture protrusions in the image, especially the outlines of axially slender objects.

[0073] Edge Difference Enhancement:

[0074] Perform edge enhancement processing (such as Sobel operator + non-maximum suppression) on the disparity image D and compare it with the original left view Superimpose and calculate image difference map , and perform threshold enhancement processing to obtain an enhanced image to further highlight the edge of the filamentous foreign body;

[0075] Directional profile fitting and feature extraction:

[0076] Extract continuous elongated boundary areas in the enhanced image and calculate their contour edge point sets;

[0077] Use sub-pixel direction fitting algorithm to calculate the main direction vector, length and center coordinates of the foreign body;

[0078] The spatial orientation properties of the foreign object are obtained, including the position coordinates (x, y), direction angle θ, and length L.

[0079] A further technical improvement of the present invention is that the method for constructing a source tracing path by the mapping module comprises the following steps:

[0080] Z1. Regional process information analysis:

[0081] The mapping module first extracts the "region number" information from the structure-enhanced image recognition data and queries the production process information corresponding to the number in the multi-field mapping table, including:

[0082] Type of process (such as surface cleaning, lamination, exposure, dispensing, curing);

[0083] The corresponding process section sequence number;

[0084] Unique identifiers of key equipment used in the current process (e.g., equipment ID);

[0085] The process parameter template ID corresponding to the process;

[0086] Z2. Production equipment parameter binding:

[0087] The module mapping module calls the log data interface in the on-site manufacturing execution system (MES) or equipment control platform (such as PLC, SCADA) based on the equipment ID and process segment time interval obtained in step Z1 to extract the following equipment operating parameter information:

[0088] Sensor parameters such as temperature, humidity, pressure, bonding force, exposure power, and operating current during the sampling period;

[0089] Equipment operation status code, fault code, and alarm event records;

[0090] Configuration parameters such as the current device software version and execution program number;

[0091] Z3. Historical defect record retrieval:

[0092] The mapping module further retrieves historical records related to the current inspection area from the defect history database, including:

[0093] Identification results of whether there are similar foreign matter types in several past batches with the same equipment, process section, and area number;

[0094] Is there a high-incidence period of defects corresponding to parameter fluctuations in this process section?

[0095] Whether there is local abnormal aggregation caused by process switching, equipment maintenance, or raw material batch changes;

[0096] Z4. Source tracing path generation and output structure:

[0097] The mapping module ultimately generates a source traceability path record based on the results of the above three steps Z2, Z3, and Z4. The record is output in a structured data format and contains at least the following fields:

[0098] Detection area number;

[0099] Foreign body type, location, and recognition confidence;

[0100] Corresponding processing technology name, process number, and processing time window;

[0101] Device ID used, device status and parameter snapshot;

[0102] Historical anomaly correlation scoring and traceability labels.

[0103] This source traceability path is used in subsequent analysis and processing modules to perform operations such as defect classification, root cause identification, process optimization suggestions, and equipment maintenance warnings.

[0104] A further technical improvement of the present invention is that: in each monitoring sub-area During the corresponding production stage, the system collects key process parameters related to the corresponding stage in real time, including temperature (unit: degrees Celsius, sampling frequency 1 Hz); pressure (unit: Newton); exposure power (unit: milliwatts per square centimeter); glue layer thickness (unit: micrometer); action duration (unit: milliseconds); specifically, the system samples each monitoring sub-area at a fixed time interval The key process parameters of the process are calculated and a sliding window mechanism is used to take the latest T time step data each time to form a multi-dimensional process parameter sequence. ;

[0105] And the system organizes the key process parameters into a multi-dimensional process parameter sequence in the form of a T*n matrix in chronological order. ;

[0106] The system integrates a time series prediction model based on long short-term memory neural network in the mapping module;

[0107] When the system is working, the receiving monitoring sub-area Multidimensional process parameter sequence for the latest time period The predicted probability of defect occurrence is obtained through the output of the time series prediction model ;

[0108] And the system will predict the probability Compared with the preset risk threshold, if the predicted probability If the risk is greater than the threshold, the corresponding monitoring sub-area The prediction results are judged as potential high-risk areas;

[0109] When a monitoring sub-area Once a potential high-risk area is identified, the system sends a dispatch instruction to the recognition and detection module to improve the image acquisition quality of the area. This includes simultaneously activating the lateral viewing angle detection subunit, the adjustable optical axis deflection subunit, and the parallax enhancement recognition subunit to achieve collaborative imaging under multiple angles, multiple parallaxes, and multiple lighting conditions, enhancing the visibility and structural resolution of foreign objects in the area.

[0110] Moreover, each of the above sub-units is a physically independent module in the system structure, which realizes parallel triggering and image data synchronization through a unified scheduling controller, and integrates their output results according to the timestamp alignment mechanism.

[0111] A further technical improvement of the present invention is that the mapping module is used to monitor the sub-area For nodes, construct a directed weighted graph structure G = (V, E), where:

[0112] , represents all monitoring sub-areas;

[0113] , represents the process dependency edge between regions, where , indicating that from the monitoring sub-area Monitoring sub-area The intensity of dependence;

[0114] In this embodiment, Obtained through historical defect co-occurrence frequency statistics, process engineering expert experience configuration, or foreign matter concentration trend modeling based on image recognition module output;

[0115] For each downstream area , calculate its cumulative risk score , expressed as follows:

[0116] ;

[0117] Where, is the predicted probability of defect occurrence; is the regional dependence intensity; To monitor the sub-area From the time it was identified as a high-risk area to the current assessment time delay; is the time series attenuation coefficient, and its specific value can be adjusted according to the time sensitivity of the actual process and the characteristics of historical failure data;

[0118] If the cumulative risk score If the value is greater than the preset judgment threshold, the downstream monitoring sub-area Marked as a secondary risk area and pushed to the identification and detection module, triggering the preset policy level upgrade.

[0119] Compared with the prior art, the present invention has the following beneficial effects:

[0120] By introducing a multidimensional process parameter sequence modeling mechanism into the mapping module and integrating deep time series prediction models such as long short-term memory (LSTM) networks, this method can predict the likelihood of future foreign material defects based on the changing trends of parameters such as temperature, pressure, and exposure power in each monitored sub-area at different process stages. This mechanism not only supports real-time scoring and risk threshold warnings but also adjusts the acquisition level of the inspection unit based on the prediction results, enabling proactive upgrades in image acquisition and improving defect detection rates.

[0121] Furthermore, the present invention further constructs a process dependency graph, treating each monitored sub-region of the display panel as a graph node. The system then defines dependency edges and weights between regions, combining historical defect co-occurrence patterns with process sequence information. When high-risk signs appear in an upstream region, the system identifies that node as the source of risk, propagates the risk score along the dependency edges to related regions, and dynamically adjusts image acquisition strategies and analysis accuracy based on the score results, forming a risk response system for chain-based defect prediction and prevention.

[0122] Furthermore, combining the structurally enhanced image recognition data output by the recognition and detection module, the mapping module in this invention further constructs a multi-field mapping table consisting of "area number - process parameters - equipment information - timestamp," ensuring that every identified defect can be traced back to a specific equipment number and process stage. By adding defect risk prediction fields and transmission path information to the mapping table, abnormal behavior can be recorded, analyzed, and responsibility assigned, effectively supporting subsequent production optimization, quality control, and equipment maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0123] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0124] Figure 1 This is the system logic diagram of the present invention. DETAILED DESCRIPTION

[0125] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0126] Example 1

[0127] See also Figure 1 As shown, the present invention provides a foreign body detection and visual recognition system for display panel production, including a monitoring area division module, an identification and detection module, a mapping module and an analysis and processing module;

[0128] The monitoring area division module is based on the display panel structure diagram and process flow diagram, and divides the display panel into multiple monitoring sub-areas with physical boundaries and functional distinctions according to its hierarchical structure (such as TFT substrate area, color film area, bonding area, and packaging glue edge area);

[0129] Each monitoring sub-area is assigned a unique area number and includes the area coordinate range, structure label (such as glass / OCA / frame), and process stage label (such as exposure / lamination / packaging);

[0130] The area numbering is synchronized with the rolling code of the production line, and the image frame and the physical panel position consistency are achieved through feature point alignment algorithms (such as image edge detection and dot target recognition);

[0131] When speed fluctuations, camera acquisition delays, and other issues occur during production line operation, causing the corresponding positions of the monitoring sub-areas in the image to shift, the system calculates the drift value based on the structural points identified in the image frame and dynamically corrects the area number to achieve self-correction of the number.

[0132] The method for the system to dynamically correct the area number includes the following steps:

[0133] S1. Determine whether there is a risk of number misalignment:

[0134] The system calculates the time difference between the acquisition time of the image frame and the time when the region number is bound;

[0135] If the theoretical production line movement distance corresponding to the time difference exceeds the preset area length threshold (for example, exceeds the length of a sub-area), it is preliminarily determined that the current image may have a number misalignment and the correction process is initiated;

[0136] S2. Extract structural feature points in the image:

[0137] Perform edge detection and feature point extraction on the current image frame to identify key structural points in the image, such as display border lines, positioning corners, fitting frame boundaries, and other parts with strong structural stability and identifiability, as a reference for determining the actual position of the image;

[0138] S3. Match with standard structure template:

[0139] In this embodiment, the standard structure template extracts the structural pattern of each monitoring sub-area from the panel design file (such as a CAD drawing, a Gerber file, or a process flow chart) to form a standard geometric outline, and the standard structure template includes at least the elements shown in Table 1 below:

[0140] Table 1

[0141]

[0142] The system extracts standard points from the standard structure template corresponding to the corresponding area number, compares them with the key structure points extracted from the image, and uses boundary contour matching, feature point pairing or image registration methods (such as affine transformation fitting) to calculate the offset between the key structure points and the standard point positions to obtain the image drift distance and direction;

[0143] S4. Determine whether the number needs to be corrected:

[0144] The system sets an offset tolerance threshold. If the actual position of the image structure point deviates from the template position by more than the tolerance threshold, the current image number is considered inconsistent with the actual area. Based on the direction and distance of the offset, the system calculates how many numbers should be adjusted forward or backward;

[0145] S5. Execution number correction:

[0146] Based on the offset judgment result, the system replaces the original number of the current image frame with the corrected area number, and records the correction information in the number correction table, including the original number, new number, drift distance, correction time, etc.

[0147] S6. Closed-loop feedback and subsequent processing:

[0148] The correction information is fed back to the monitoring area division module. If the number offset occurs in multiple consecutive image frames, the system will output an alarm message to indicate that there is a synchronization anomaly;

[0149] At the same time, the corrected number is used to bind the subsequent identification data to ensure that the identification results are correctly attributed to their actual structural areas;

[0150] For example, taking the fifth sub-area in the upper left corner of the panel as an example, its structure template contains the following data:

[0151] Area code: N5;

[0152] Structural feature outline: rectangular window + bevel positioning line;

[0153] Standard feature point coordinates: {(120,45),(260,45),(260,160),(120,160)};

[0154] Image direction angle: 0°;

[0155] Resolution: 3.5µm / pixel;

[0156] When the border edge point in the detected image frame deviates from the coordinate range by more than ±8 pixels, the system automatically triggers the number correction;

[0157] The recognition and detection module uses the microstructural differences of the display panel as input. By sensing the physical structural properties, process state characteristics, and potential foreign object types of the test area, it dynamically configures the image acquisition method and imaging parameters to achieve the optimal response effect in the image of potential foreign objects in the monitoring sub-area.

[0158] The recognition and detection module obtains a set of local image features through multi-view image fusion, multi-angle illumination scanning, and disparity map construction, performs enhancement processing, and outputs a set of standardized structure-enhanced image recognition data;

[0159] The local image feature set is the identifiable visual response features such as brightness, contrast, edge, texture, etc. of the corresponding area in the image formed by foreign matter (such as particles, filaments, bubbles, residual glue, etc.) on the surface or structural layer of the display panel under specific image acquisition conditions;

[0160] Structural enhancement image recognition data includes: enhanced images of foreign objects in the corresponding monitoring sub-area, the spatial location of the foreign object within the area, the foreign object morphology classification results, image enhancement method labels, recognition confidence scores, and area number information. This ensures that foreign objects with different attachment methods, different structural hierarchical positions, and different orientation characteristics can be recognized with high confidence under a unified enhanced recognition process. That is, when the system identifies foreign objects, it can provide high-precision judgment results based on image data, that is, the system's recognition process has high accuracy and reliability. Specifically:

[0161] Accuracy: The recognition results are highly consistent with the actual situation, and the probability of misidentification and missed recognition is low.

[0162] Reliability: The system has a high degree of trust in the results. A high confidence score generally means that the prediction results of the recognition model are very stable and not easily affected by noise or errors.

[0163] Structural enhancement image recognition data improves the visibility of foreign object images while retaining the location information of foreign objects in the spatial structure of the display panel, which is used to support subsequent defect classification, spatial coordinate calculation and process traceability analysis.

[0164] The recognition and detection module includes a lateral viewing angle detection subunit, an adjustable optical axis deflection subunit, and a parallax enhancement recognition subunit;

[0165] The above sub-units are deployed simultaneously in the system structure, and the system can selectively activate one or more sub-units to perform image acquisition and enhancement processing based on the predicted results of the monitoring sub-area structure, process status or foreign body type;

[0166] The lateral viewing angle detection subunit is used to identify non-frontal visible defects in the edge area or structural parts of the display panel due to poor bonding, loose structure or micro-foreign matter adhesion, including residual bonding colloid, edge film material warping, structural component bulging, etc.

[0167] Specifically, the lateral viewing angle detection subunit sets a plurality of non-vertically mounted industrial cameras in a specific monitoring sub-area of ​​the display panel, and a fixed tilt angle is formed between the optical axis of each camera and the normal direction of the panel, and the tilt angle range is 30° to 60°, preferably 45°;

[0168] The layout locations include but are not limited to:

[0169] The four border areas of the panel (used to detect the edge sealing structure);

[0170] Polarizer lamination edge (for detecting laminated foreign matter);

[0171] FPC bonding area (used to detect foreign matter at the side outlet);

[0172] At the same time, a main downward-looking camera is placed directly above the panel, forming a joint acquisition array with the above-mentioned side cameras;

[0173] The workflow of the lateral view detection subunit includes:

[0174] Image acquisition: When the target area enters the inspection station, the system synchronously triggers the top-view camera and all side-view cameras to acquire images, obtaining top-view images and side-view images respectively, ensuring that the image timestamps are consistent;

[0175] Spatial registration: Perform affine transformation and angle correction on the top view image and each side view image. Specifically, this includes: extracting the image edge point set and performing preliminary registration with the boundary features in the structural template; solving the affine matrix using the edge corner points and performing coordinate remapping; and uniformly transforming the top view image and each side view image into a regional image in a standard reference coordinate system.

[0176] Image fusion and contrast enhancement: Perform pixel-weighted fusion on the remapped images from each perspective to obtain a fused image and enhance edge structure signals;

[0177] Detect the grayscale variability of local areas in the fused image to generate a preliminary heat map of suspected foreign bodies;

[0178] Edge defect feature extraction: In the fused image, select the preset corner area ROI and perform: Canny edge extraction; local texture directional variation detection based on Gabor filter; connection analysis and morphological filtering to extract abnormal contour areas;

[0179] Defect determination: Based on the contour integrity, texture interference index, and structural symmetry index, a classification model is constructed to classify and determine the following defect types that may exist in the corner area. In this embodiment, the results of the determination are shown in Table 2:

[0180] Table 2

[0181]

[0182] The adjustable optical axis deflection subunit is used to improve the visibility of slightly tilted, weakly adhered, or low-reflectivity foreign objects on the display panel surface. It is particularly suitable for detecting lightly adhered particles, tilted adhesive residues, transparent foreign objects, and other defects that are difficult to image under conventional vertical incident light. The hardware structure includes but is not limited to the following:

[0183] Controllable deflection component: preferably using a MEMS micromirror array, a dual-axis servo pan / tilt, or an electric mirror mechanism to precisely control the incident direction of light, with a deflection angle range of ±30°;

[0184] Directional light source: emits a focusable and controllable directional lighting beam, coupled with a deflection component to ensure local lighting at any angle;

[0185] Optical axis control driver: with high-precision angle control capability, the angle adjustment step accuracy is preferably 5°;

[0186] Incident angle database: stores the recommended incident angle range and preferred angle value associated with each monitoring sub-area number, which is obtained based on historical training data, structural morphology and material reflection characteristics;

[0187] Its workflow includes:

[0188] Incident angle configuration acquisition: When a certain area number i enters the current detection station, the system reads the incident angle setting value set corresponding to the area from the incident angle database ;

[0189] Angle scanning acquisition: Control the controllable deflection component to set the incident angle one by one to the collection Various angles in ;

[0190] Each angle set , that is, synchronously trigger the industrial camera to collect images , until all angle images are collected and the image sequence is obtained ={ };

[0191] Image enhancement processing: image sequence Perform local contrast enhancement processing (such as CLAHE, gradient boosting) on ​​each frame of the image and evaluate the response intensity of the foreign body target area ;

[0192] If there is a strongest response image , then the frame image is selected as the final enhanced image;

[0193] If the difference in the response intensity of n frames is less than the preset difference threshold, the image stacking and fusion processing is performed to generate the final enhanced image ;

[0194] Foreign body recognition execution: the image or Input the preset defect recognition network to classify and locate weak reflectors, low-contrast particles, and semi-transparent areas, and output the defect type, location, area, and reflection intensity labels;

[0195] The defect recognition network uses a convolutional neural network (CNN) model trained through deep learning. This model is supervised using a dataset of display panel defect images collected by the system. The dataset covers the following types of foreign matter:

[0196] Adherent particulate foreign matter (glass chips, dust, impurities);

[0197] Translucent spots of residual glue;

[0198] Micro-reflectors visible at oblique incidence;

[0199] Traces of membrane material falling off;

[0200] Each image in the dataset contains manually annotated information such as the foreign body location box, category label, defect outline, and visual score;

[0201] The parallax enhancement recognition subunit is used to identify slender foreign objects (such as filaments, hair, and thin fibers) on the display panel surface that are highly directional and axially parallel to the main line of sight. This solves the detection blind spot problem of such foreign objects in vertical top-down imaging, where they have weak reflections, fuzzy shapes, and unclear boundaries. Specifically:

[0202] The parallax enhancement recognition subunit consists of two industrial cameras positioned on either side of the target monitoring subarea of ​​the display panel. Both cameras are tilted ±15° to the left and right relative to the panel normal, ensuring their line of sight is horizontal. Both cameras are equipped with synchronized triggering devices to ensure consistent image acquisition timing and eliminate motion interference. The geometric relationship between the cameras is determined through a post-installation calibration process, including their respective intrinsic parameter matrices and distortion coefficients (for distortion correction), as well as their rotational and translational relationships (for geometric calibration).

[0203] Its workflow includes synchronous image acquisition, disparity map construction, edge difference enhancement, and directional contour fitting and feature extraction;

[0204] Synchronous image acquisition:

[0205] When the detection station reaches the target area, the two cameras belonging to the parallax enhancement recognition subunit respectively capture the image of the current monitoring sub-area, which is recorded as the left view. and right view , and the parallax enhancement recognition subunit for the left view and right view Perform image distortion correction and geometric calibration to obtain the left view and right view Paired images of

[0206] In this embodiment, image distortion correction refers to acquiring images through a standard calibration plate after the camera is installed to obtain distortion coefficients and intrinsic parameters for correcting imaging distortion caused by the lens. Specifically, the parallax enhancement recognition subunit uses the calibration results of each camera (including the intrinsic parameter matrix and distortion coefficients) to correct the left view. and right view Remap the pixel positions of the distorted areas in the image to eliminate the radial and tangential distortion caused by the lens and generate a distortion-corrected image.

[0207] Geometric calibration refers to the system's calibration of the left view based on the installation angle and relative position (i.e., external parameters) between the cameras. and right view Perform geometric calibration and registration transformations to align the two images in spatial structure, ensuring that the pixels of the targets in the images are completely aligned in the vertical direction when calculating the disparity, and only retaining the disparity difference in the horizontal direction, thus forming a standard image pairing relationship;

[0208] Specifically, the image distortion correction and geometric calibration and registration methods are any mature methods in the prior art, such as Zhang Zhengyou camera calibration method, OpenCV stereo image correction, and other publicly available and mature engineering methods;

[0209] Disparity map construction:

[0210] Left view and right view Execute stereo matching algorithms based on block matching or semi-global matching;

[0211] Output a disparity image D, where the grayscale value of each pixel in the disparity image D represents the horizontal offset of that pixel between the two images. The disparity image D can highlight structural features with slight height variations or texture protrusions in the image, especially the outlines of axially slender objects.

[0212] Edge Difference Enhancement:

[0213] Perform edge enhancement processing (such as Sobel operator + non-maximum suppression) on the disparity image D and compare it with the original left view Superimpose and calculate image difference map , and perform threshold enhancement processing to obtain an enhanced image to further highlight the edge of the filamentous foreign body;

[0214] Directional profile fitting and feature extraction:

[0215] Extract continuous elongated boundary areas in the enhanced image and calculate their contour edge point sets;

[0216] Use sub-pixel direction fitting algorithms (such as least squares straight line fitting or ellipse major axis extraction) to calculate the main direction vector, length, and center coordinates of the foreign body;

[0217] Obtain the spatial orientation properties of the foreign body, including position coordinates (x, y), direction angle θ, and length L;

[0218] The mapping module is used to identify the association and binding between the structure-enhanced image recognition data output by the detection module and the corresponding processing technology, production equipment and historical parameters, and to build a traceability path for the source of foreign body recognition results;

[0219] The mapping module receives the recognition result including the area number information, and based on the mapping relationship between the area number and the production process code, queries and extracts the processing steps, processing equipment numbers, equipment operation time periods and corresponding key process parameters corresponding to the monitoring sub-area;

[0220] Key process parameters include, but are not limited to, coating thickness (μm), exposure energy (mJ / cm²), lamination pressure (kgf), UV irradiation intensity (W / cm²), lamination temperature (°C), and ambient temperature and humidity during processing (%RH, °C). These process parameters are provided by the sensor network or MES system during the production process and are matched and bound to the monitoring sub-area number using timestamps and panel numbers.

[0221] The mapping module further establishes a multi-field mapping table containing "monitoring sub-area number - processing equipment number - process parameter set - timestamp" to accurately map each defect data point in the foreign body identification results to its potential processing source, providing input basis for equipment responsibility confirmation, process parameter traceability, and subsequent analysis and processing;

[0222] The method for constructing a source traceability path by a mapping module includes the following steps:

[0223] Z1. Regional process information analysis:

[0224] The mapping module first extracts the "region number" information from the structure-enhanced image recognition data and queries the production process information corresponding to the number in the multi-field mapping table, including:

[0225] Type of process (such as surface cleaning, lamination, exposure, dispensing, curing);

[0226] The corresponding process section sequence number;

[0227] Unique identifiers of key equipment used in the current process (e.g., equipment ID);

[0228] The process parameter template ID corresponding to the process;

[0229] Z2. Production equipment parameter binding:

[0230] The module mapping module calls the log data interface in the on-site manufacturing execution system (MES) or equipment control platform (such as PLC, SCADA) based on the equipment ID and process segment time interval obtained in step Z1 to extract the following equipment operating parameter information:

[0231] Sensor parameters such as temperature, humidity, pressure, bonding force, exposure power, and operating current during the sampling period;

[0232] Equipment operation status code, fault code, and alarm event records;

[0233] Configuration parameters such as the current device software version and execution program number;

[0234] Z3. Historical defect record retrieval:

[0235] The mapping module further retrieves historical records related to the current inspection area from the defect history database, including:

[0236] Identification results of whether there are similar foreign matter types in several past batches with the same equipment, process section, and area number;

[0237] Is there a high-incidence period of defects corresponding to parameter fluctuations in this process section?

[0238] Whether there is local abnormal aggregation caused by process switching, equipment maintenance, or raw material batch changes;

[0239] Z4. Source tracing path generation and output structure:

[0240] The mapping module ultimately generates a source traceability path record based on the results of the above three steps Z2, Z3, and Z4. The record is output in a structured data format and contains at least the following fields:

[0241] Detection area number;

[0242] Foreign object type, location, and recognition confidence (from structure-enhanced image recognition data);

[0243] Corresponding processing technology name, process number, and processing time window;

[0244] Device ID used, device status and parameter snapshot;

[0245] Historical anomaly correlation scoring and traceability labels.

[0246] This source traceability path is used in subsequent analysis and processing modules to perform operations such as defect classification, root cause identification, process optimization suggestions, and equipment maintenance warnings;

[0247] The analysis and processing module is used to integrate the structural enhancement image recognition data output by the recognition and detection module with the processing traceability data generated by the mapping module to perform defect risk assessment, quality anomaly trend analysis, and cross-region defect correlation judgment. It includes the following processing functions:

[0248] Early warning mechanism for potential defects under process fluctuations:

[0249] The analysis and processing module continuously receives process parameter data streams from the mapping module and sets historical baseline values ​​and deviation tolerance thresholds for each parameter type. If a monitoring sub-area shows abnormal fluctuations in the corresponding processing parameters (such as a continuous temperature deviation or a short-term sudden increase in exposure energy), even though no obvious foreign matter has been identified, the system marks the area as a "potential abnormal area" and outputs a warning indicator to prompt subsequent stricter inspections or shutdown verification.

[0250] Defect common cause identification mechanism:

[0251] For foreign objects identified in multiple different areas or continuous panels, the analysis and processing module uses methods such as cluster analysis, structural similarity matching, and equipment path overlap analysis to determine whether they may be from the same source of contamination or equipment anomalies from the same batch. If multiple conditions such as location proximity, similar morphological characteristics, and consistent processing equipment are met, the system will merge and label the above defect data as "common cause defects" and locate the suspected responsible equipment or material batch;

[0252] Number mismatch correction and data closed-loop feedback mechanism:

[0253] Under high-speed production line conditions, if image acquisition delays or transmission errors cause misalignment between image and area numbers, the analysis and processing module uses image feature point re-identification, number sliding window comparison, and timeline consistency verification mechanisms to detect number anomalies in real time and automatically correct the numbering of incorrectly bound image data to ensure the unique matching relationship between image, area, equipment, and parameters. At the same time, the numbering correction information is fed back to the monitoring area division module, forming an endogenous data closed loop in the system.

[0254] The final output of the analysis and processing module includes: the risk level label of each defect, the judgment result of whether it belongs to the common cause defect cluster, the identification signal of whether an early warning is triggered, and the corrected number-bound data for subsequent production optimization, equipment inspection and quality judgment.

[0255] Example 2

[0256] A foreign body detection and visual recognition system for display panel production, comprising a monitoring area division module, an identification and detection module, a mapping module, and an analysis and processing module;

[0257] The monitoring area division module is based on the display panel structure diagram and process flow diagram, and divides the display panel into multiple monitoring sub-areas with physical boundaries and functional distinctions according to its hierarchical structure (such as TFT substrate area, color film area, bonding area, and packaging glue edge area);

[0258] Each monitoring sub-area is assigned a unique area number and includes the area coordinate range, structure label (such as glass / OCA / frame), and process stage label (such as exposure / lamination / packaging);

[0259] The area numbering is synchronized with the rolling code of the production line, and the image frame and the physical panel position consistency are achieved through feature point alignment algorithms (such as image edge detection and dot target recognition);

[0260] When speed fluctuations, camera acquisition delays, and other issues occur during production line operation, causing the corresponding positions of the monitoring sub-areas in the image to shift, the system calculates the drift value based on the structural points identified in the image frame and dynamically corrects the area number to achieve self-correction of the number.

[0261] The recognition and detection module uses the microstructural differences of the display panel as input. By sensing the physical structural properties, process state characteristics, and potential foreign object types of the test area, it dynamically configures the image acquisition method and imaging parameters to achieve the optimal response effect in the image of potential foreign objects in the monitoring sub-area.

[0262] The recognition and detection module enhances the foreign object image signal through multi-view image fusion, multi-angle illumination scanning, disparity map construction, and edge feature fitting, and outputs a set of standardized structure-enhanced image recognition data. The structure-enhanced image recognition data includes: the foreign object enhanced image of the corresponding monitoring sub-area, the spatial position of the foreign object within the area, the foreign object morphology classification result, the image enhancement method label, the recognition confidence score, and the area number information. This ensures that foreign objects with different attachment methods, different structural hierarchical positions, and different orientation characteristics can be recognized with high confidence under a unified enhanced recognition process.

[0263] Structural enhancement image recognition data improves the visibility of foreign object images while retaining the location information of foreign objects in the spatial structure of the display panel, which is used to support subsequent defect classification, spatial coordinate calculation and process traceability analysis.

[0264] The mapping module is used to identify the association and binding between the structure-enhanced image recognition data output by the detection module and the corresponding processing technology, production equipment and historical parameters, and to build a traceability path for the source of foreign body recognition results;

[0265] The mapping module receives the recognition result including the area number information, and based on the mapping relationship between the area number and the production process code, queries and extracts the processing steps, processing equipment numbers, equipment operation time periods and corresponding key process parameters corresponding to the monitoring sub-area;

[0266] The mapping module further establishes a multi-field mapping table containing "monitoring sub-area number - processing equipment number - process parameter set - timestamp" to accurately map each defect data point in the foreign body identification results to its potential processing source, providing input basis for equipment responsibility confirmation, process parameter traceability, and subsequent analysis and processing;

[0267] The analysis and processing module is used to integrate the structure-enhanced image recognition data output by the recognition and detection module with the processing traceability data generated by the mapping module to perform defect risk assessment, quality anomaly trend analysis, and cross-regional defect correlation judgment;

[0268] The final output of the analysis and processing module includes: the risk level label of each defect, the judgment result of whether it belongs to the common cause defect cluster, the identification signal of whether an early warning is triggered, and the corrected number-bound data for subsequent production optimization, equipment inspection and quality judgment.

[0269] Compared with Example 1, Example 2 has During the corresponding production stages (such as bonding, exposure, and packaging), the system collects key process parameters related to the corresponding stages in real time, including temperature (unit: degrees Celsius, sampling frequency 1 Hz); pressure (unit: Newton); exposure power (unit: milliwatts per square centimeter); and glue layer thickness (unit: micrometer).

[0270] Action duration (unit: milliseconds); specifically, the system samples each monitoring sub-area at a fixed time interval The key process parameters of the process are calculated and a sliding window mechanism is used to take the latest T time step data each time to form a multi-dimensional process parameter sequence. ;

[0271] And the system organizes the key process parameters into a multi-dimensional process parameter sequence in the form of a T*n matrix in chronological order. , which is expressed as follows:

[0272] ;

[0273] Where, represents the value of the i-th key process parameter at the j-th time point, T is the length of the sliding time window; n is the number of key process parameters; each column is the time series of a key process parameter; each row is the sampling value of all parameters corresponding to a certain moment;

[0274] The system integrates a time series prediction model based on long short-term memory neural network in the mapping module;

[0275] The time series prediction model uses historical production data, where the input is the process parameter sequence within a certain time window, and the output is the label of whether there is a defect identified by the recognition and detection module after the end of the window. During training, the input is the historical multi-dimensional process parameter sequence , and outputs the defect label confirmed by the recognition and detection module after the target time period; the defect label is reverse-labeled by the foreign body recognition result obtained from the structure-enhanced image recognition data;

[0276] The training set of the time series prediction model contains at least 10,000 samples, each of which corresponds to a specific monitoring sub-area. ;

[0277] When the system is working, the receiving monitoring sub-area Multidimensional process parameter sequence for the latest time period The predicted probability of defect occurrence is obtained through the output of the time series prediction model ;

[0278] And the system will predict the probability Compared with the preset risk threshold, if the predicted probability If the risk is greater than the threshold, the corresponding monitoring sub-area The prediction results are judged as potential high-risk areas;

[0279] When a monitoring sub-area Once a potential high-risk area is identified, the system sends a dispatch instruction to the recognition and detection module to improve the image acquisition quality of the area. This includes simultaneously activating the lateral viewing angle detection subunit, the adjustable optical axis deflection subunit, and the parallax enhancement recognition subunit to achieve collaborative imaging under multiple angles, multiple parallaxes, and multiple lighting conditions, enhancing the visibility and structural resolution of foreign objects in the area.

[0280] Moreover, each of the above sub-units is a physically independent module in the system structure, which realizes parallel triggering and image data synchronization through a unified scheduling controller, and integrates their output results according to the timestamp alignment mechanism.

[0281] Example 3

[0282] Compared with Example 1, the mapping module in Example 3 is based on monitoring sub-areas For nodes, construct a directed weighted graph structure G = (V, E), where:

[0283] , represents all monitoring sub-areas;

[0284] , represents the process dependency edge between regions, where , indicating that from the monitoring sub-area Monitoring sub-area The intensity of dependence;

[0285] In this embodiment, Obtained through historical defect co-occurrence frequency statistics, process engineering expert experience configuration, or foreign matter concentration trend modeling based on image recognition module output;

[0286] For example, if the upstream monitoring sub-area To monitor the sub-areas for the bonding process For the packaging process, experience shows that the former defect is easy to affect the latter, so set ;

[0287] When a monitoring sub-area If the detection module determines that the area is a potential high-risk area, the system will traverse all process-dependent edges in the directed weighted graph, that is, traverse the directed weighted graph structure G = (V, E), and monitor the sub-area with the current potential high-risk area. For all outgoing edges from the starting point ( → ), used to identify potentially affected downstream areas .

[0288] Example 4

[0289] Compared with Example 2, the mapping module in Example 4 is used to monitor the sub-area For nodes, construct a directed weighted graph structure G = (V, E), where:

[0290] , represents all monitoring sub-areas;

[0291] , represents the process dependency edge between regions, where , indicating that from the monitoring sub-area Monitoring sub-area The intensity of dependence;

[0292] In this embodiment, Obtained through historical defect co-occurrence frequency statistics, process engineering expert experience configuration, or foreign matter concentration trend modeling based on image recognition module output;

[0293] When a monitoring sub-area If the detection module determines that the area is a potential high-risk area, the system will traverse all process-dependent edges in the directed weighted graph, that is, traverse the directed weighted graph structure G = (V, E), and monitor the sub-area with the current potential high-risk area. For all outgoing edges from the starting point ( → ), used to identify potentially affected downstream areas ;

[0294] For each downstream area , calculate its cumulative risk score , expressed as follows:

[0295] ;

[0296] Where, is the predicted probability of defect occurrence; is the regional dependence intensity; To monitor the sub-area From the time it was identified as a high-risk area to the current assessment time delay; is the time series attenuation coefficient, and its specific value can be adjusted according to the time sensitivity of the actual process and the characteristics of historical failure data;

[0297] If the cumulative risk score If the value is greater than the preset judgment threshold, the downstream monitoring sub-area Marked as a secondary risk area and pushed to the identification and detection module, triggering the preset policy level upgrade.

[0298] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field according to actual conditions.

[0299] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. 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 content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A foreign body detection and visual recognition system for display panel production, characterized in that: include: The monitoring area division module divides the display panel into multiple monitoring sub-areas based on the display panel structure diagram and process flow diagram, and assigns a unique area number to each monitoring sub-area; The recognition and detection module performs image detection on each monitoring sub-area, obtains and enhances the local image feature set, and outputs a set of standardized structure-enhanced image recognition data; The mapping module associates the structure-enhanced image recognition data with the corresponding processing technology, production equipment, and historical parameters, constructs a traceability path for the source of foreign body recognition results, and establishes a multi-field mapping table of "monitoring sub-area number - processing equipment number - process parameter set - timestamp"; The analysis and processing module integrates structure-enhanced image recognition data and source traceability paths to perform defect risk assessment, quality anomaly trend analysis, and cross-regional defect correlation judgment.

2. The foreign body detection and visual recognition system for display panel production according to claim 1, characterized in that: When the recognition and detection module performs image detection, if the corresponding position of the monitoring sub-area in the image is offset, the system dynamically corrects the area number, including the following steps: S1, the system calculates the time difference between the acquisition time of the image frame and the region number binding time; If the theoretical production line movement distance corresponding to the time difference exceeds the preset area length threshold, it is determined that the current image has a number misalignment; S2. Perform edge detection and feature point extraction on the current image frame to identify key structural points in the image frame; S3, extracting standard points from the standard structure template corresponding to the corresponding area number, and comparing them with the key structure points extracted from the image frame, calculating the offset between the key structure points and the standard points, and obtaining the image drift distance and direction; S4. Determine whether the offset exceeds a preset offset tolerance threshold. If so, it is considered that the area number of the current monitoring sub-area is misaligned. S5. According to the offset determination result, the original number of the current image frame is replaced with the corrected region number.

3. The foreign body detection and visual recognition system for display panel production according to claim 1, characterized in that: The recognition and detection module includes a lateral viewing angle detection subunit, an adjustable optical axis deflection subunit and a parallax enhancement recognition subunit; The lateral viewing angle detection subunit, the adjustable optical axis deflection subunit and the parallax enhancement recognition subunit are deployed simultaneously in the system structure, and the system selectively activates one or more subunits to perform image acquisition and enhancement processing.

4. The foreign body detection and visual recognition system for display panel production according to claim 3, characterized in that: The side view angle detection subunit synchronously captures the top view image and the side view image of the display panel by deploying multiple non-vertically mounted industrial cameras, and performs image registration and enhancement processing on the captured top view image and the side view image; Through the processed top view image and side view image, the lateral viewing angle detection subunit identifies and extracts defects in the edge area or structural parts of the display panel.

5. The foreign body detection and visual recognition system for display panel production according to claim 4, characterized in that: The adjustable optical axis deflection subunit is used to adjust the optical axis direction of the image acquisition device during the display panel inspection process so that light is irradiated to the monitoring sub-area to be tested at different incident angles, thereby enhancing the imaging contrast of foreign object edges, transparent contours and multi-layer interfaces.

6. The foreign body detection and visual recognition system for display panel production according to claim 5, characterized in that: The parallax enhancement recognition subunit is used to construct a parallax signal spectrum between different perspectives by synchronously collecting image data from multiple fixed perspectives within the monitoring sub-area of ​​the display panel, identify image features that change morphologically due to changes in perspective, and extract the displacement response of foreign objects between different structural levels or transparent materials.

7. The foreign body detection and visual recognition system for display panel production according to claim 1, characterized in that: The method for constructing a source tracing path by the mapping module includes the following steps: Z1. Regional process information analysis: The mapping module first extracts the "region number" information from the structure-enhanced image recognition data and searches the multi-field mapping table for the production process information corresponding to the number, including: Type of process (such as surface cleaning, lamination, exposure, dispensing, curing); The corresponding process section sequence number; Unique identifiers of key equipment used in the current process (e.g., equipment ID); The process parameter template ID corresponding to the process; Z2. Production equipment parameter binding: The mapping module calls the log data interface in the on-site manufacturing execution system or equipment control platform based on the equipment ID and process time interval obtained in step Z1 to extract equipment operating parameter information; Z3. Historical defect record retrieval: The mapping module further retrieves historical records related to the current inspection area from the defect history database, including: Identification results of whether there are similar foreign matter types in several past batches with the same equipment, process section, and area number; Is there a high-incidence period of defects corresponding to parameter fluctuations in this process section? Whether there is local abnormal aggregation caused by process switching, equipment maintenance, or raw material batch changes; Z4. Source tracing path generation and output structure: The mapping module ultimately generates a source traceability path record based on the results of the above three steps Z2, Z3, and Z4. The record is output in a structured data format and contains at least the following fields: Detection area number; Foreign body type, location, and recognition confidence; Corresponding processing technology name, process number, and processing time window; Device ID used, device status and parameter snapshot; Historical anomaly correlation scoring and traceability labels.

8. The foreign body detection and visual recognition system for display panel production according to claim 1, characterized in that: In each monitoring sub-area During the corresponding production stage, the system collects key process parameters related to the corresponding stage in real time; And the system organizes the key process parameters into a multi-dimensional process parameter sequence in the form of a T*n matrix in chronological order. , which is expressed as follows: ； Where, represents the value of the i-th key process parameter at the j-th time point, T is the length of the sliding time window; n is the number of key process parameters; The system integrates a time series prediction model based on long short-term memory neural network in the mapping module; When the system is working, the receiving monitoring sub-area Multidimensional process parameter sequence for the latest time period The predicted probability of defect occurrence is obtained through the output of the time series prediction model ; And the system will predict the probability Compared with the preset risk threshold, if the predicted probability If the risk is greater than the threshold, the corresponding monitoring sub-area The prediction results determine it as a potential high-risk area.

9. The foreign body detection and visual recognition system for display panel production according to claim 8, characterized in that: Mapping modules to monitor sub-areas For nodes, construct a directed weighted graph structure G = (V, E), where: , represents all monitoring sub-areas; , represents the process dependency edge between regions, where , indicating that from the monitoring sub-area Monitoring sub-area The intensity of dependence; When a monitoring sub-area If the identification and detection module determines that the area is a potential high-risk area, the system traverses all process-dependent edges in the directed weighted graph; For each downstream area , calculate its cumulative risk score , expressed as follows: ； Where, is the predicted probability of defect occurrence; is the regional dependence intensity; To monitor the sub-area From the time it was identified as a high-risk area to the current assessment time delay; is the time series attenuation coefficient, and its specific value can be adjusted according to the time sensitivity of the actual process and the characteristics of historical failure data; If the cumulative risk score If the value is greater than the preset judgment threshold, the downstream monitoring sub-area Marked as secondary risk area.

Images