Full-color Micro LED micro-display module packaging detection method, device and equipment

Abstract

The invention relates to the field of packaging detection, in particular to a full-color Micro LED micro display module packaging detection method, device and equipment, and the method comprises the following steps: carrying out pixel-level thermal induction on a micro display module to be detected, collecting global pixel thermal induction self-luminous response parameters, carrying out instantaneous light response analysis, and generating an instantaneous light response curvature; multi-time-point response trend analysis is carried out based on the instantaneous light response curvature, global pixel thermal disturbance coupling evolution is carried out, and a pixel thermal disturbance response map is constructed; performing inter-pixel multi-dimensional response difference calculation and abnormal deviation region positioning according to the pixel thermal disturbance response map, and extracting an abnormal deviation display module region; performing spatial arrangement inversion on the abnormal deviation display module area to generate actual pixel point coordinates of the abnormal area; according to the method, error compensation is carried out on the offset pixels through precise and efficient huge transfer error analysis, and the packaging precision and reliability control level of the micro-display module are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of display screen packaging detection, and in particular to a full-color Micro LED micro display module packaging detection method, device and equipment. BACKGROUND

[0002] With the rapid development of new generation display technology, Micro LED is gradually becoming an important development direction in the future display field due to its high brightness, high contrast, low power consumption, long service life and other significant advantages. Especially in the application scenarios of augmented reality (AR), virtual reality (VR), wearable devices, high-end vehicle display and ultra-high-definition video wall, Micro LED display technology shows great market potential. Among them, the performance and yield of full-color Micro LED micro display module as the core display unit directly affect the display effect and cost control of the whole system, and become the focus of research in the industry and academia.

[0003] In the manufacturing process of Micro LED micro display module, mass transfer technology is a key link to realize large-scale and high-density pixel arrangement. This process needs to accurately transfer thousands of red, green and blue (RGB) micro LED chips from the transfer carrier to the driving substrate under micron-level precision to complete the assembly of color pixels. Compared with the traditional packaging method, mass transfer technology has significant advantages in efficiency and resolution, but also faces many technical challenges, especially the transfer yield and precision control problems are increasingly prominent.

[0004] In the mass transfer process, due to the small size and high density of the chips, any error may cause the chips to be offset, missing, flipped or damaged, which not only affects the consistency of light emission and color accuracy, but also may even cause the module to fail. In addition, environmental factors, equipment precision, unstable process control, etc. may also cause a series of hidden defects. These defects are often difficult to be discovered in time by traditional detection methods before packaging is completed, thereby causing production yield to decrease, cost to rise and product reliability to decrease. At present, the commonly used detection methods mainly include optical imaging detection, electrical performance testing and manual visual inspection, etc. These methods have certain effect in detecting simple defects, but when facing high-density and multi-pixel full-color micro display modules, their detection efficiency and accuracy are obviously insufficient. Especially in the initial stage of mass transfer, the displacement error of small chips, foreign matter pollution, empty position and other micro defects are difficult to be accurately identified in real time, which cannot meet the needs of high-precision and high-reliability packaging quality control. SUMMARY

[0005] The present application is to solve the above technical problems, and proposes a full-color Micro LED micro display module packaging detection method, device and equipment to solve at least one of the above technical problems.

[0006] To achieve the above object, the present application provides a full-color Micro LED micro display module packaging detection method, comprising the following steps: Step S1: pixel-level thermal induction is performed on the micro display module to be detected, global pixel thermal induction spontaneous light emission response parameters are collected, transient light response analysis is performed, and transient light response curvature is generated; Step S2: based on the transient light response curvature, multi-time point response trend analysis is performed, and global pixel thermal disturbance coupling evolution is performed to construct a pixel point thermal disturbance response map; Step S3: pixel-to-pixel multi-dimensional response difference calculation and abnormal deviation area positioning are performed according to the pixel point thermal disturbance response map, and an abnormal deviation display module area is extracted; Step S4: spatial arrangement inversion is performed on the abnormal deviation display module area, and accurate pixel point positioning calculation is performed to generate actual pixel point coordinates of the abnormal area; Step S5: a preset pixel point design arrangement blueprint is obtained, pixel transfer error calculation and adaptive pixel transfer error compensation are performed on the actual pixel point coordinates, and transfer error compensation parameters are generated; Step S6: based on the transfer error compensation parameters, a large amount of transfer process parameter optimization simulation is performed, dynamic packaging control is performed, and an intelligent packaging detection optimization model is constructed.

[0007] In the present specification, a full-color Micro LED micro display module packaging detection device is provided for performing the method as described above, comprising: A thermal induction module is configured to perform pixel-level thermal induction on the micro display module to be detected, collect global pixel thermal induction spontaneous light emission response parameters, perform transient light response analysis, and generate transient light response curvature; A response trend module is configured to perform multi-time point response trend analysis based on the transient light response curvature, and perform global pixel thermal disturbance coupling evolution to construct a pixel point thermal disturbance response map; An abnormal area positioning module is configured to perform pixel-to-pixel multi-dimensional response difference calculation and abnormal deviation area positioning according to the pixel point thermal disturbance response map, and extract an abnormal deviation display module area; A pixel point calculation module is configured to perform spatial arrangement inversion on the abnormal deviation display module area, and perform accurate pixel point positioning calculation to generate actual pixel point coordinates of the abnormal area; A transfer error compensation module is configured to obtain a preset pixel point design arrangement blueprint, perform pixel transfer error calculation and adaptive pixel transfer error compensation on the actual pixel point coordinates, and generate transfer error compensation parameters; The transfer process optimization module is used for performing mass transfer process parameter optimization simulation based on transfer error compensation parameters and dynamic package control, and constructing an intelligent package detection optimization model.

[0008] The application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the full-color MicroLED micro display module package detection method according to any one of the preceding embodiments when executing the computer program.

[0009] The beneficial effects of the present application are as follows: by applying thermal disturbance at the pixel level, the spontaneous light response changes of each pixel point in the micro display module are induced, which can realize non-contact and non-invasive dynamic detection without damaging the module structure. The generation of instantaneous light response curvature can sensitively reflect the response characteristic differences of the pixels under the action of thermal disturbance, providing high-resolution and high-sensitivity response data basis for subsequent global analysis, and significantly improving the detection accuracy and data integrity. Through time series analysis of multi-time point response data, the response trend and change law of the pixels under the action of thermal disturbance are identified, which can reveal the nonlinear evolution behavior of the thermal response between the pixels. At the same time, the construction of the thermal disturbance response atlas can realize the dynamic global modeling of the entire pixel array, so that the system can intuitively analyze the coupling region and the pixel response mode, improve the identification and modeling ability of complex thermal interference coupling phenomenon, and provide theoretical support for subsequent accurate positioning and compensation. Using the constructed response atlas, the pixel response behavior is extracted and analyzed through multi-dimensional features such as response delay, decay rate and intensity drift, the response differences between each pixel are accurately quantified, and abnormal regions are automatically identified through data clustering and deviation threshold setting. This processing method greatly improves the intelligence and automation level of detection, can effectively eliminate incidental errors, quickly lock the deviated regions in the mass transfer, and is helpful to intervene and repair problems in advance, reducing yield loss. Through the spatial geometric relationship inversion and coordinate analysis of the abnormal region, the positioning calculation at the pixel level is realized, which ensures that the system can identify the specific pixel position where the deviation or failure exists. The high-precision coordinate extraction not only improves the accuracy of subsequent compensation calculation, but also provides intuitive guidance for repair technology and repeated detection. The method takes into account the spatial analysis and error tracking capability, which is an important technical link to improve the transfer precision control and promote the industrialization of mass transfer. By comparing the actual detection results with the design blueprint, the error quantification of the deviated pixels is realized through the error vector calculation model, and an adaptive compensation strategy is constructed combined with statistical characteristics and historical deviation mode. The generation of compensation parameters provides a basis for the equipment to realize dynamic adjustment and closed-loop control, significantly reduces the cumulative deviation caused by repeated errors, and improves the stability and consistency of the mass transfer process, thereby greatly enhancing the packaging precision and reliability control level. Based on the foregoing compensation parameters, a process simulation and optimization system for mass transfer is constructed, and parameter iterative optimization is realized by simulating the influence of different process parameters on the transfer precision. At the same time, a dynamic packaging control mechanism is introduced to adjust the key process factors in real time during the actual packaging process, effectively control the transfer deviation and reduce the accumulation risk of system error. The finally constructed intelligent packaging detection and optimization model can realize end-to-end process control, and lay a technical foundation for large-scale high-precision Micro LED packaging. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 A flowchart of the steps of a method of the present application; Figure 2A detailed implementation step flowchart for step S1 is shown in the following figure; Figure 3 A detailed implementation step flowchart for step S2 is shown in the following figure; Figure 4 A detailed implementation step flowchart for step S3 is shown in the following figure. DETAILED DESCRIPTION

[0011] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0012] The present application provides a full-color Micro LED micro display module packaging detection method, device and equipment. The execution subject of the method, device and equipment includes but is not limited to the following: mechanical equipment, data processing platform, cloud server node, network upload device, etc. which can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: audio image management system, information management system, cloud data management system at least one.

[0013] Please refer to Figures 1 to 4 The present application provides a full-color Micro LED micro display module packaging detection method, which includes the following steps: Step S1: pixel-level thermal induction is performed on the micro display module to be detected, global pixel thermal induction spontaneous light emission response parameters are collected, transient light response analysis is performed, and transient light response curvature is generated; Step S2: based on the transient light response curvature, multi-time point response trend analysis is performed, and global pixel thermal disturbance coupling evolution is performed to construct a pixel point thermal disturbance response map; Step S3: pixel-to-pixel multi-dimensional response difference calculation and abnormal deviation area positioning are performed according to the pixel point thermal disturbance response map, and an abnormal deviation display module area is extracted; Step S4: spatial arrangement inversion is performed on the abnormal deviation display module area, and accurate pixel point positioning calculation is performed to generate actual pixel point coordinates of the abnormal area; Step S5: a preset pixel point design arrangement blueprint is obtained, pixel transfer error calculation and adaptive pixel transfer error compensation are performed on the actual pixel point coordinates, and a transfer error compensation parameter is generated; Step S6: based on the transfer error compensation parameter, a large amount of transfer process parameter optimization simulation is performed, dynamic packaging control is performed, and an intelligent packaging detection optimization model is constructed.

[0014] In the embodiment of the present application, please refer to Figure 1A flowchart of the steps of the full-color Micro LED micro display module packaging detection method of the present application is shown in the present example. The steps of the full-color Micro LED micro display module packaging detection method include: Step S1: pixel-level thermal induction is performed on the micro display module to be detected, global pixel thermal induction spontaneous light emission response parameters are collected, transient light response analysis is performed, and transient light response curvature is generated; In this embodiment, the micro display module is placed on a thermal field generator. The thermal field generator can be an electric heating element or a laser heater, which can accurately control the output of heat. The temperature sensor is used to monitor the temperature change of the module surface in real time, ensuring that the thermal field uniformly covers the entire module. In this process, the initial temperature is recorded for subsequent analysis. The intensity of the excitation signal should be set to about 1 W / cm² per pixel to ensure that each pixel can respond fully. Once the thermal induction starts, a high-resolution camera and a spectrometer are used to collect the spontaneous light emission response parameters of the global pixel thermal induction in real time. The camera should be set in a position parallel to the module to ensure that it can capture the light emission of each pixel. The data acquisition frequency is set to 10 Hz to monitor the light intensity and changes of each pixel in real time. The light intensity data of each pixel is extracted through image processing software to form a detailed light response database. After obtaining the pixel spontaneous light emission response parameters, transient light response analysis is performed. The goal of this process is to evaluate the light emission characteristics of each pixel under thermal induction. Experimental parameters are set, such as setting the analysis window of the transient light response to 500 milliseconds to capture the dynamic changes of light emission in the initial stage of thermal excitation. Signal processing techniques such as Fast Fourier Transform (FFT) are used to perform spectral analysis on the collected light intensity data, thereby identifying the frequency characteristics of light emission changes. The transient light response curvature is generated based on the above analysis results. The generation of curvature can be achieved by calculating the light intensity change rate of each pixel within a time window. Experimental parameters are set, such as setting the time step for curvature calculation to 100 milliseconds to obtain a higher time resolution. Through numerical analysis methods, the second derivative of the light intensity change is calculated to obtain the value of the light response curvature. This curvature will provide information about the sensitivity and stability of the thermal response of each pixel.

[0015] Step S2: based on the transient light response curvature, multi-time point response trend analysis is performed, and global pixel thermal disturbance coupling evolution is performed to construct a pixel point thermal disturbance response map; In this embodiment, the light response curvature values at multiple time points are extracted from the instantaneous light response curvature data. The time range for analysis is set to the first 500 milliseconds after thermal excitation, and 5 key time points (such as 0 ms, 100 ms, 200 ms, 300 ms, and 400 ms) are selected. At this time, the curvature values at these time points are sorted using data processing software (such as MATLAB or Excel) to form time series data. In this process, ensure the accuracy and integrity of the data to prevent data deviation caused by external interference. To improve the reliability of the analysis, data smoothing processing can be performed, such as using the moving average method, to reduce the impact of noise on the results. The extracted multi-time point curvature data is statistically analyzed. Linear regression analysis or polynomial fitting method is used to identify the response trend. The fitting accuracy requirement is set to more than 95% to ensure the effectiveness and reliability of the analysis results. In this process, by drawing the graph of curvature change over time, observing the shape and slope change of the curve, the degree of influence of thermal disturbance on pixel response is identified. If the curvature changes significantly at certain time points, special attention should be paid to these moments for in-depth analysis of the physical mechanism. After completing the multi-time point response trend analysis, the analysis of the global pixel thermal disturbance coupling evolution is carried out. The goal is to explore the mutual influence of each pixel under thermal disturbance conditions. A coupling relationship model between pixels is established, and the coupling strength threshold is set to 0.5 to screen out pixel pairs with significant coupling relationship. By calculating the correlation between the response curvature of each pixel and the response curvature of its adjacent pixels, the pixel group that responds together under thermal disturbance is identified. Cross-correlation function analysis is used to help understand the coupling effect between pixels and record the response characteristics of key pixels in the coupling evolution process. Based on the above analysis results, a pixel thermal disturbance response map is constructed. This map will show the response characteristics and coupling relationship of different pixels under thermal disturbance. The resolution of the map is set to 1 mm per pixel to ensure that the map has sufficient detail. Visualization software (such as Origin or Matplotlib) will be used to generate heat maps or three-dimensional graphics to make the response characteristics of different regions more intuitive. By color coding different light response intensities and coupling degrees, it is easy to quickly identify areas with good performance and potential faults. Through this series of steps, we can comprehensively analyze the thermal disturbance response characteristics of the full-color Micro LED micro display module and construct a detailed response map. This method will significantly improve the detection accuracy and efficiency of the micro display module and provide important basis for subsequent quality control and performance optimization.

[0016] Step S3: Perform multi-dimensional response difference calculation and abnormal deviation area positioning between pixels according to the pixel thermal disturbance response map, and extract the abnormal deviation display module area; In this embodiment, relevant data needs to be extracted from the pixel thermal disturbance response map. These data include the luminous intensity, response time, and curvature of each pixel. Set the experimental parameters, for example, select 3 key dimensions for response difference calculation: luminous intensity, response time, and curvature. Organize these features into a multi-dimensional array for subsequent analysis. Use data processing software (such as MATLAB or Python) to standardize the extracted pixel feature data. This process ensures that data in different dimensions is within the same range to avoid analysis bias due to different dimensions. After standardization, each characteristic value will be subtracted from its mean and divided by the standard deviation to form a standard score. After completing the standardization, multi-dimensional response difference calculation between pixels is performed. Set the experimental parameters, for example, set the response difference threshold to 1.5 times the standard deviation to facilitate the identification of significant response differences. Use Euclidean distance or Mahalanobis distance methods to calculate the response difference between each pair of adjacent pixels. Specifically, for each pixel, calculate the difference in selected dimensions with adjacent pixels and form a difference matrix. Each element in the difference matrix represents the response difference between two pixels in multiple dimensions. Once the response difference matrix is obtained, the next step is to locate the abnormal deviation area. Set the experimental parameters, for example, set the identification threshold of abnormal deviation to 20% of the maximum value in the response difference matrix. By setting conditional judgments, identify pixels with difference values exceeding the threshold and mark them as abnormal responses. Use image processing techniques to generate a heat map or marker map of the abnormal area. Through this visualization method, you can intuitively identify which areas have performance problems. Use clustering algorithms (such as K-means) to further analyze the distribution characteristics of abnormal pixels and identify possible fault areas. Extract the specific fault area in the display module from the identified abnormal deviation area. Set the experimental parameters, for example, set the minimum area threshold for extracting the area to 5 pixels to avoid false positives. By using connected component analysis method, adjacent abnormal pixel areas are merged to form a complete fault area. Output the location information of the abnormal deviation area, including coordinates and area parameters, and generate a detailed report. These information will provide the basis for subsequent fault analysis and quality control, so that manufacturers can make targeted repairs or adjustments.

[0017] Step S4: Space arrangement inversion is performed on the abnormal deviation display module area, and precise pixel point positioning calculation is performed to generate the actual pixel point coordinates of the abnormal area. In this embodiment, a preliminary spatial model needs to be generated based on the previously identified abnormal deviation area. Set the experimental parameters, such as setting the boundary of the abnormal area to 5 pixels to ensure the accuracy and integrity of the model. Use image processing software (such as MATLAB or OpenCV) to extract the contour of the abnormal area to form a polygon or closed curve to represent the spatial distribution of the abnormal area. Convert the extracted contour to coordinates to ensure consistency with the coordinate system of the actual display module. Set the experimental parameters, such as setting the origin of the coordinate system as the upper left corner of the display module in pixels to facilitate subsequent calculation and analysis. After generating the spatial model of the abnormal area, perform accurate pixel point positioning calculation. Set the experimental parameters, such as setting the size of each pixel to 0.1mm to ensure the accuracy of the calculation. Use image processing technology to calculate the specific coordinates of each pixel based on the contour of the abnormal area. Use the grid division method to divide the abnormal area into several small blocks, each containing several pixels. Calculate the center point of each small block to obtain the pixel point coordinates within the abnormal area. Specifically, use the centroid calculation formula to find the centroid coordinates of each small block and use them as the representative coordinates of the region. After completing the pixel point positioning, organize all the pixel coordinates of the abnormal area. Set the experimental parameters, such as setting the data output format to CSV file to facilitate subsequent processing and analysis. Use data processing software to organize the coordinates, status (normal or abnormal) and other information of each pixel point into a table. In this process, ensure the accuracy of the coordinate data and perform a second check on the coordinates of each abnormal pixel to avoid errors in the data processing process affecting subsequent analysis. Use data visualization tools to mark the coordinates of abnormal pixels on the image to visually display the distribution of the abnormal area. Based on the organized data, generate the actual pixel point coordinates of the abnormal area. These coordinates will be used for subsequent fault analysis and repair decisions. Set the experimental parameters, such as setting the coordinate precision to two decimal places to ensure the accuracy of the positioning. Output the final coordinate file and generate a detailed analysis report containing the distribution of the abnormal area, the coordinates of each abnormal pixel and its status. These information will provide important basis for subsequent quality control, fault troubleshooting and performance optimization, enabling manufacturers to make targeted repairs and improvements.

[0018] Step S5: Obtain a preset pixel point design arrangement blueprint, and perform pixel transfer error calculation and adaptive pixel transfer error compensation on the actual pixel point coordinates, to generate a transfer error compensation parameter; In this embodiment, the preset pixel design arrangement blueprint needs to be obtained from the design team. This blueprint is usually displayed in CAD or other graphic formats, and the design position, size, and spacing of each pixel must be clearly visible. Set the experimental parameters, such as setting the resolution of the blueprint to 300 DPI to ensure the accuracy of the details. After obtaining the blueprint, use image processing software (such as AutoCAD or Adobe Illustrator) to convert it into a digital format. By extracting the design coordinates of each pixel in the blueprint, a standard pixel coordinate list is generated. These coordinates will serve as the basis for subsequent error calculations. After completing the acquisition of the design blueprint, the transfer error calculation of the actual pixel coordinates and the design coordinates is carried out. Set the experimental parameters, such as setting the error calculation threshold to ±0.1 mm to ensure that the error is within an acceptable range. In specific implementation, first compare the actual pixel coordinates with the corresponding coordinates in the design blueprint one by one. Use the Euclidean distance formula to calculate the distance between each actual pixel and its design coordinates, thereby obtaining the pixel transfer error. Record the error value of each pixel and calculate the average error and maximum error of the entire module to facilitate subsequent analysis. Once the transfer error is calculated, the next step is to perform adaptive pixel transfer error compensation. The goal of this process is to achieve matching with the design blueprint by adjusting the actual pixel position. Set the experimental parameters, such as setting the compensation adjustment range to ±0.2 mm to ensure the effectiveness and reasonableness of the compensation. In the implementation of compensation, first calculate the coordinates that need to be adjusted according to the transfer error of each pixel. Specifically, for each actual pixel, the difference between its coordinates and the design coordinates is used as the compensation parameter. Then, using the calculated compensation parameters, adjust the actual position of each pixel to ensure that it matches the design blueprint. Based on the above compensation results, generate the transfer error compensation parameters. These parameters will be used in subsequent production and calibration processes, set the experimental parameters, such as setting the output format of the compensation parameters to CSV file to facilitate subsequent processing and application. When generating compensation parameters, ensure that the compensation value of each pixel is clearly visible, and relevant error statistical information is attached. Output the final compensation report, which includes the original coordinates, design coordinates, transfer error, and final compensation coordinates of each pixel. These information will provide important basis for subsequent quality control and production adjustment, so that the micro display module can achieve the expected performance standard in actual application.

[0019] Step S6: Based on the transfer error compensation parameters, the mass transfer process parameter optimization simulation is carried out, and the dynamic packaging control is carried out to construct an intelligent packaging detection optimization model.

[0020] In this embodiment, the optimization simulation of the mass transfer process parameters needs to be performed based on the previously generated transfer error compensation parameters. Set the experimental parameters, for example, set the simulated transfer speed to 25 mm / s to reflect the transfer speed in actual production. In this process, multiple key process parameters need to be selected for optimization, including transfer pressure, transfer time, and temperature, etc. Use simulation software (such as ANSYS or COMSOL) to establish a physical model of the packaging process. The model should include factors such as material flow, temperature change, and pressure distribution during the transfer process to ensure the accuracy of the simulation. By setting boundary conditions and initial conditions, ensure that the model can truly reflect the actual packaging process. Once the model is established, perform optimization simulation of the process parameters. The goal of this process is to find the best combination of transfer process parameters to minimize the transfer error. Set the experimental parameters, for example, set the optimization objective function to the minimum value of the transfer error, and use optimization methods such as genetic algorithm or particle swarm optimization algorithm to solve. During the simulation process, gradually adjust parameters such as transfer pressure, transfer time, and temperature, and observe their effect on the transfer error. Through multiple iterations, record the transfer error after each optimization, and draw a response surface map to visually display the effect of different parameter combinations on the transfer effect. After completing the process parameter optimization, the next step is to implement dynamic packaging control. The goal of this process is to monitor and adjust process parameters in real time during the packaging process to ensure high precision and stability of the transfer. Set the experimental parameters, for example, set the monitoring frequency to 10 times per second to ensure real-time data. When implementing dynamic control, use sensors (such as temperature sensors and pressure sensors) to monitor environmental parameters in real time during the packaging process. Feed the real-time data back to the control system, and use closed-loop control strategies to automatically adjust the transfer parameters to respond to environmental changes or material property fluctuations. Based on the optimization and control results described above, construct an intelligent packaging detection optimization model. This model will integrate the optimized process parameters, dynamic control strategies, and real-time monitoring data to form a comprehensive packaging quality control system. Set the experimental parameters, for example, set the model update frequency to every time the packaging is completed, so as to timely reflect the process improvement. When building the model, use machine learning algorithms (such as support vector machines or decision trees) to train historical packaging data to identify key factors affecting packaging quality. Through real-time analysis of new data, the model will be able to predict and optimize potential problems in the packaging process.

[0021] In this embodiment, refer to Figure 2 For the detailed implementation step flowchart of step S1, in this embodiment, the detailed implementation steps of step S1 include: Define different thermal field excitation signals to induce pixels of the micro display module at the pixel level, and collect global pixel thermal induction spontaneous light emission response parameters; Based on the self-luminous response parameter, initial luminous intensity of each RGB sub-pixel point is calculated, and initial luminous intensity of pixel points under different thermal fields is extracted; According to the initial luminous intensity of the pixel points, time sequence light intensity change tracking is performed to obtain a dynamic light intensity change curve of each pixel point; Based on the pixel-level thermal-induced self-luminous response log, a thermal field loading first timestamp and an initial luminous timestamp are calculated; According to the thermal field loading first timestamp and the initial luminous timestamp, a thermal excitation response delay time of each sub-pixel point is calculated; According to the thermal excitation response delay time, transient light response analysis of the dynamic light intensity change curve under thermal field disturbance is performed to generate a transient light response curvature.

[0022] In this embodiment, a controllable thermal field excitation system is established for precise pixel-level thermal excitation of the Micro LED micro display module. The excitation signal can adopt pulse thermal excitation (Pulse Thermal Excitation) and step thermal excitation (Step Thermal Excitation) two forms, respectively for short-time response and steady-state response test. The thermal field excitation system is generally composed of a MEMS micro-heater matrix, and each heating unit corresponds to a pixel or sub-pixel point of the module. The amplitude of the excitation signal is set in the range of 40-120℃, and the excitation time is controlled between 0.1ms-10ms. In order to avoid the error caused by non-uniform heating, a multi-point temperature control feedback loop should be used, and real-time monitoring is performed through infrared thermal imaging (such as FLIR A655sc) to ensure the uniformity of thermal excitation. At the same time, during the thermal field excitation process, a high-speed acquisition system is used to monitor the real-time light emission of the Micro LED module. Using a high-speed camera (such as Photron FASTCAM series) with a resolution of at least 1920x1080 and a frame rate of more than 1000 fps, combined with a microlens array, the light intensity of each sub-pixel can be collected. After completing the light response acquisition under thermal excitation, the next step is to quantitatively calculate the initial light intensity of each RGB sub-pixel. The "initial light intensity" here is defined as the light intensity value when the first significant light signal appears after the start of thermal excitation, and the light intensity threshold with a signal-to-noise ratio greater than 10 is usually selected. Using image segmentation algorithm (such as sub-pixel positioning model based on YOLO or U-Net) to segment the collected image frame by frame, the coordinates and light emission area of each sub-pixel are extracted. Converting the gray value of each sub-pixel at different time points to physical light intensity value needs to be calibrated. The calibration process is performed by calibrating the spectrum radiation of the LED module before the experiment, and using an integrating sphere and a spectrometer (such as CAS 140CT) to jointly determine the output intensity curve of the RGB sub-pixel under different driving conditions. By matching the current pixel gray map under thermal excitation with the calibration curve through interpolation method, the initial light intensity of each sub-pixel under different thermal fields can be accurately calculated.

[0023] With the sub-pixel positioning information obtained in the previous step, the extraction region is fixed, and the light intensity of the collected high-speed image sequence is counted frame by frame. The light intensity change curve of each pixel point will show the intensity rising, stable or falling stage with time growth, and the specific response form depends on its packaging structure, heat diffusion performance and material response characteristics. In order to ensure the high fidelity of the dynamic change curve, the sliding window filtering (window size is set to 3-5 frames) is used to denoise the light intensity data, and the complete curve is reconstructed by using polynomial fitting or spline interpolation. The time axis of the data is taken as the reference zero point (t=0) of the thermal excitation trigger, so as to ensure the time sequence alignment of the response process of different pixels. The final output result is a three-dimensional matrix: the ID of each pixel point × time stamp × light intensity value. This matrix can be further used for local anomaly detection, such as judging which pixels have too low response amplitude or too slow response at a specific time point, prompting packaging defects or electro-thermal coupling problems. Two key time points are extracted from the response log: the first time stamp of the thermal field loading (T0) and the initial light-emitting time stamp of the pixel (T1). T0 is defined as the time point when the thermal field heating module starts to act on the sample, which is usually recorded by the thermal field control system. If it is an automated test platform, this time stamp can be automatically embedded in the image sequence metadata by the control software. T1 is defined as the time point when each pixel point first reaches significant light emission under thermal excitation, which is specifically defined as the time when the gray value of the pixel first exceeds three times (3σ) of the background noise. A dynamic threshold algorithm based on edge detection can be used to search for this transition point on each pixel curve. The difference ΔT = T1-T0 is the response delay of the pixel thermal excitation, which represents the time required for the thermal energy to be conducted from the input end to the effective light-emitting area and to induce electroluminescence. This delay time has obvious differences for different packaging areas and different sub-pixel color types, and is an important parameter for subsequent thermal coupling characteristic analysis. The response delay time of all RGB sub-pixels is systematically analyzed. The ΔT value is counted according to the sub-pixel category (R, G, B) respectively, and the thermal excitation response delay distribution map is obtained. Further division into mean response time, maximum response time, response time standard deviation and other statistical indicators can reflect the consistency and thermal diffusion uniformity of the module packaging. This analysis result not only helps to detect the response delay abnormalities caused by thermal resistance or material unevenness, but also can be used to predict the light response speed of the pixel in actual use, providing a reference for subsequent drive optimization and module calibration. Under ideal packaging conditions, the thermal excitation response difference of RGB three types of pixels should not exceed ±0.5 ms; if the response delay of some blue light pixels reaches more than 2 ms, micro welding defects or thermal diffusion layer damage problems need to be considered.

[0024] High-order feature extraction is performed on the dynamic light intensity curve, and a curvature of instantaneous light response (CILR) is calculated. The parameter can be used to represent the light-emitting sensitivity and change rate of the pixel under thermal excitation. The curvature of instantaneous light response is realized by calculating the second derivative, and the formula is: κ(t)=(y^'' (t)) / ([1+(y^' (t))^2 ]^(3 / 2) ); wherein y(t) is a function of the change of light intensity with time, y^' (t) and y^'' (t) are first and second derivatives, which can be approximated by the five-point difference method. The curvature curve is calculated for all pixel curves, and the maximum curvature point and the corresponding time are extracted to identify response mutations, wave delay and other phenomena. Generally, high curvature corresponds to the mutation response after thermal excitation, and low curvature indicates slow response or light output saturation. By constructing a curvature heat map of the full module, it can be directly presented which areas have good package uniformity and response sensitivity, and which areas have hidden dangers.

[0025] In this embodiment, referring to Figure 3 For the detailed implementation step flow diagram of step S2, in this embodiment, the detailed implementation steps of step S2 include: Identify the positive and negative response values of each pixel point based on the curvature of instantaneous light response; Perform multi-time point response trend analysis on the positive and negative response values to obtain the thermal disturbance response characteristics of each pixel point, including the leading type, the lagging type and the abnormal type; Define the response type based on the thermal disturbance response characteristics to obtain the response type label of each pixel point; Perform full-time sequence light intensity trend mining on the dynamic light intensity change curve to obtain the full-time sequence light intensity trend characteristics of each pixel point; Label each pixel point based on the response type label, and display the global pixel thermal disturbance coupling evolution of the display module according to the full-time sequence light intensity trend characteristics, and construct a pixel point thermal disturbance response map.

[0026] In this embodiment, on the basis of the instantaneous light response curvature (CILR) of each pixel calculated in the previous step, this step needs to further extract positive and negative response values for judging the response directionality and fluctuation trend of the pixel to thermal disturbance. The positive response value represents the acceleration of the rising trend of the light intensity of the pixel under the action of thermal excitation (positive curvature), and the negative response value indicates that the pixel response appears to decline or lag (negative curvature). In order to extract the positive and negative response characteristics, zero-crossing point detection and extreme point identification are required for the curvature curve. In the experiment, the numerical difference method is used to perform first-order difference on each curvature curve, and a threshold (±0.01) is set to filter the pseudo-curvature caused by background noise. In addition, in order to enhance the analysis stability, a minimum response threshold (such as the maximum curvature must be higher than 0.05 to be counted as a positive response) can be set to avoid misjudgment of weak disturbance as effective response. Each pixel will finally record a "response polarity sequence", for example [+, +, -, +, -, -], representing the response directionality at multiple thermal excitation time points. This response polarity information is the basis for thermal response classification, and also reflects the linkage quality of the pixel package to thermal conduction and light output. The positive and negative response sequences of each pixel at multiple thermal excitation time points are analyzed for time trend, and the response mode on the time axis is identified. The classification is mainly defined by the following three types of response characteristics: Advanced response: continuous positive response appears in the early stage of thermal excitation (such as within the first 50% of the thermal disturbance window), representing high thermal sensitivity and rapid conduction of thermal excitation to the light-emitting area.

[0027] Delayed response: response mainly concentrates after the end of thermal excitation or after a few milliseconds of thermal input lag, indicating that the thermal conduction path is delayed, which may be caused by high thermal resistance at the package layer interface or poor micro soldering.

[0028] Abnormal response: frequent positive and negative fluctuation switching, response amplitude significantly deviating from the mean value (more than 3σ), etc. in the response sequence, which may correspond to pixel structure variation, material non-uniformity or internal cavity effect. The analysis method uses principal component analysis (PCA) to reduce the dimension of the positive and negative response time sequence, and then uses K-means clustering (k=3) for automatic clustering to classify into the above three types of response mode. In the experiment, it is recommended to select 5 equally spaced thermal excitation time points as multi-time point input, so that each pixel has a 5-dimensional response polarity vector (for example [+1, +1, -1, +1, -1], converted to numerical processing), which is further used for trend classification.

[0029] The response characteristics obtained in the previous step are attributed to a clear response type label for building subsequent thermal response maps and state tracking. The label definitions are as follows: Type-A (advance type), Type-D (lag type), Type-X (abnormal type). Supervised learning modeling is performed on the PCA dimension reduction results and the artificial preliminary classification results based on discriminant analysis (such as LDA) to realize batch automatic response labeling. The specific method is to use part of the known abnormal area or reference standard module as the training data set, extract its thermal response law, and establish a discriminant function. Each pixel finally obtains a response type label, which is attached to the time sequence record of the pixel and becomes part of the metadata of the pixel structure thermal response. Experimental statistics show that the proportion of Type-X pixels of a qualified module should be less than 0.5%; if it exceeds 2%, it can be preliminarily judged that there is a systematic packaging risk. The light intensity data of each pixel from the start of thermal excitation to the complete stability of light emission are extracted to form an intensity situation descriptor (ISD) for describing the complete response trend. This process uses a multi-scale time window analysis method to extract the following features from the curve: peak intensity (Peak, rise rate (Rise Slope), time-to-peak (Time-to-Peak), stable drift (Stable Drift), curve standard deviation and skewness, Each feature corresponds to a certain packaging or structural characteristic from a physical level, for example, "rise rate" is related to thermal conduction efficiency, and "stable drift" is related to material thermal expansion and optical coupling consistency. In data processing, a sliding window method (window length 10 frames, step 1 frame) can be used to generate a local fitting curve, and then statistical regression is used to calculate each feature value to finally form a situation description matrix containing a 6-10 dimensional feature vector. This matrix can be compared with the standard module for similarity when the module is detected, and the light emission stability level of each pixel can be quantified. The response type of each pixel is visualized in the image coordinate system, and a thermal disturbance response map (TRM) is constructed to macroscopically present the response evolution path of the entire module under thermal field excitation.

[0030] The map construction method is as follows: Arrange the module pixels according to two-dimensional coordinates; Use color coding to map the response type label (such as red for abnormal type, green for advance type, and yellow for lag type); Superimpose the principal components of the light intensity situation features of each pixel (such as the first three principal components as color depth, transparency, brightness, etc. dimensions); Optional dynamic evolution map: update the map in real time during the whole process of thermal excitation, and show the evolution process of pixel response in video mode. This response map is not only used for encapsulation detection result presentation, but also as a feedback signal to guide the encapsulation process optimization. In practice, typical defect modes include: Type-D response in edge area, local Type-X pixel group, etc., which usually correspond to poor thermal sealing or non-uniform COB binding.

[0031] In this embodiment, refer to Figure 4 For the detailed implementation step flowchart of step S3, in this embodiment, the detailed implementation steps of step S3 include: Identify the same color channel pixel points of the micro display module to be detected; According to the pixel point thermal disturbance response map, the same color channel pixel points are divided into adjacent pixels, and the multi-dimensional response difference between pixels is calculated to obtain the light response difference tensor matrix between adjacent pixels; Calculate the response deviation of each dimension of the light response difference tensor matrix, and extract the multi-dimensional response deviation value; According to the preset response difference distance threshold, the multi-dimensional response deviation value is identified for abnormal response, and the abnormal deviation area is located, and the abnormal deviation display module area is extracted.

[0032] In this embodiment, all sub-pixel points belonging to the same color channel (i.e. red R, green G or blue B) are accurately identified from the full-color Micro LED micro display module, providing a color-consistent sample basis for subsequent thermal disturbance response analysis. Since each pixel point contains RGB three-color sub-pixels, a high-resolution microscopic image acquisition device is required to obtain the complete light-emitting array image of the display module. Generally, an industrial microscopic imaging system with a resolution of more than 10 million pixels is selected, and the image is collected in a dark room to avoid environmental light interference. After image acquisition, pixel point recognition processing is performed. The image is converted from RGB color space to HSV or Lab color space to enhance color separation capability. By setting a specific color threshold, for example, red channel can select pixel points with Hue in 0°±15°, combined with Saturation and Brightness (brightness) range screening, the target color sub-pixel can be accurately extracted. Then, the recognition result is further optimized using spatial distribution rules. For example, most Micro LED modules use periodic array arrangement, and by matching the pixel arrangement template (template matching), the position error caused by slight deviation of the collection angle can be identified and corrected. Finally, a set of sub-pixel point coordinates with clear two-dimensional spatial coordinates and consistent color properties is obtained, which is used for subsequent thermal disturbance response tracking and adjacent pixel division. After completing the same color channel pixel recognition, the module is excited by applying a spatially uniform or locally controllable thermal disturbance field. Before, during and after thermal excitation, consecutive self-luminous image sequences are collected to record the luminous behavior changes of each pixel point under the action of thermal disturbance. This process can use a high-speed imaging device (such as a sCMOS camera with a frame rate of ≥1000 fps) in combination with a microscope to perform dynamic brightness acquisition, while a thermocouple array or infrared imaging records temperature changes.

[0033] The luminance change curve L(t) of each identified pixel point in the entire time window is extracted from the image sequence to obtain its thermal response behavior spectrum. To compare the responses between pixels, adjacent relationship division based on spatial coordinates is required. The most common is to use four-neighborhood (i.e. each pixel point and its four adjacent pixels above, below, left and right form a pixel pair) or eight-neighborhood division (including diagonal direction adjacent pixels).

[0034] For each adjacent pixel pair (Pi, Pj), their thermal disturbance response characteristics are calculated, such as: Peak luminance difference AL_max: represents the difference between the maximum luminance of the two pixels; Response delay difference AT: represents the time difference from the start of thermal excitation to the detection of luminance; Rising rate difference AR: the reciprocal of the time required for the light intensity to rise from 10% to 90%; Curvature difference of light intensity change ΔK: represents the acceleration difference of the brightness change in the response process. The above characteristic values are combined into a response difference vector D(i,j) = [ΔL_max, Δτ, ΔR, ΔK...], and all adjacent pixel pairs are calculated, and finally a multi-dimensional tensor matrix T containing the response differences of all pixel pairs is formed, which has the structure of NxM, where N is the number of pixel pairs, and M is the number of difference dimensions.

[0035] Subsequently, the Z-score of each pair of adjacent pixels in this dimension, i.e. the degree of deviation in the sense of standard deviation, is calculated: Z m (i,j) = (D m (i,j) - μ m ) / σ m This value can quantify the significant degree of deviation of a certain pixel pair from the average response in a certain dimension. By analogy, each pixel pair in the tensor will correspond to a normalized deviation vector Z(i,j) = [Z1, Z2, Z3,... Z m ] in each dimension.

[0036] In order to facilitate unified analysis and sorting, the comprehensive response deviation distance E(i,j) of each pixel pair can be further calculated, and the commonly used methods are Euclidean distance or weighted Manhattan distance: E(i,j) = sqrt(Σ m Z m 2 ) or E(i,j) =Σ m w m * |Z m |, where w m is the dimension weight.

[0037] The greater this comprehensive deviation distance, the more inconsistent the thermal response behavior between the pixel pair, and the possibility of process abnormality or material structure discontinuity. In order to enhance the robustness, a non-normal distribution index based on the median absolute deviation (MAD) can be introduced to replace the standard deviation, and the sensitivity control of local extreme abnormal values is enhanced. The data in the tensor matrix is normalized and abnormally distinguished, and the specific method is to extract statistical characteristics and standardize deviation scores in each dimension. For each response dimension m (such as brightness difference, response delay, etc.), the statistical mean μ m and the standard deviation σ m, which measures the natural fluctuation range of this dimension under normal circumstances. After obtaining the response deviation value E(i, j) of each pair of adjacent pixels, it is necessary to identify the pixel pairs with abnormal thermal response by comparing with the preset response difference threshold. The threshold E_th can be set based on multi-module data statistics, generally set as E_th = μE + 3σE (empirical 3σ rule) or set the optimal classification boundary through machine learning. When E(i, j) > E_th, it is determined that there is a response abnormality between the pixel pairs, indicating that at least one pixel point may have an abnormal response to thermal disturbance due to packaging problems, poor thermal interface or material non-uniformity. All abnormal pixel pairs are remapped to a two-dimensional spatial coordinate system to construct a response deviation heat map, and spatial clustering processing is performed on these points. DBSCAN (density-based clustering algorithm) or connected domain analysis algorithm can be used to divide the abnormal points into regions, so as to extract the central region of abnormal response. The distribution form, position and range of these abnormal regions provide a spatial explanation for the quality abnormality of the module, which may correspond to: thermal conduction path blockage (such as interface bubbles, incomplete colloid filling), optical non-uniformity (such as light guide film flaws), local hysteresis response caused by microcracks, The abnormal area can be visualized and output in an image superposition manner, and combined with the pixel number and position information, it can be used for rapid identification of abnormal products on the production line or guidance for manufacturing process optimization.

[0038] In this embodiment, step S4 includes the following steps: extracting the light response difference tensor matrix of the abnormal deviation display module area; spatial arrangement inversion of the light response difference tensor matrix according to a preset pixel point design arrangement blueprint to obtain a current pixel point arrangement diagram; physically sensing and modeling the neural network of the pixel point thermal disturbance response map to obtain the mapping rule of the thermal disturbance response and the spatial coordinates; based on the mapping rule, performing accurate pixel point positioning calculation on the current pixel point arrangement diagram to generate the actual pixel point coordinates of the abnormal area.

[0039] In this embodiment, according to the clustering or connected domain analysis result of the previous step, a list of pixel pair or pixel point IDs judged to be abnormal is obtained. In two-dimensional space, these IDs can be mapped to a specific coordinate point set (such as P1(x1, y1), P2(x2, y2)...). Then, from the originally constructed response difference tensor T ∈ nxm (where n is the total number of pixel pairs, and m is the response difference dimension) extract the subset tensor T_abn ∈ kxm, k is the number of abnormal pixel pairs. This tensor T abn preserves the original response difference value of each dimension (brightness difference, delay, curvature, etc.), which can be used for subsequent atlas modeling, inversion, and physical fitting operations. In order to enhance the stability of the spatial structure, it is recommended to add an additional circle of boundary pixel points when extracting the abnormal region tensor, forming a "buffer zone" to avoid missing edge information. In addition, to enhance the analysis accuracy, the T abn can be reduced in dimension based on principal component analysis (PCA) or independent component analysis (ICA), and the most physically meaningful response principal axis can be extracted for subsequent modeling tasks. Load the standard arrangement blueprint data structure of the module, which can be derived from the design CAD diagram or pixel circuit mapping table, in the form of a two-dimensional or three-dimensional coordinate index matrix. The blueprint can be represented as B(x, y, c), where x and y are pixel positions, and c is color channel information (such as R=1, G=2, B=3). By pixel recognition number and the calibration parameters between the image coordinate system and the blueprint coordinate system (usually obtained from the microscope scale and macro displacement system, with a calibration accuracy of within ±0.5 μm), the actual image acquisition coordinates are converted back to the standard arrangement coordinate system. The inversion method can use perspective transformation, affine mapping, or interpolation calibration functions, such as: If the three-point calibration method is used, an affine matrix M is established for conversion; If it is a nonlinear local distortion, B-spline or local weighted regression (LOWESS) can be used for position fitting.

[0040] Through this inversion step, each pixel pair or pixel point in the abnormal tensor is restored to its actual position on the design diagram, forming a physical coordinate arrangement map of the abnormal region. This map not only provides intuitive visual presentation, but also can be used as feedback for subsequent modeling and detection results to the packaging manufacturing end. Through data-driven methods, a nonlinear physical mapping relationship between the thermal disturbance response behavior and the pixel spatial position is established to achieve precise positioning, especially in the presence of local optical interference, packaging distortion, and other non-ideal situations while maintaining high robustness. The response curve L(t) of each pixel point (a sequence of brightness changes over time) is used as the input signal, and its position (x, y) in the arrangement map is used as the output target to train a neural network model. The model is essentially a physical perception type time series regression network, which can use the following structure: Input layer: thermal response time series (500 frames can be taken from 0-300 ms); Feature extraction module: CNN can be used to process local time slices, or LSTM can be used to extract time series dynamic features; Physical constraint module: introduce material thermal conductivity, packaging structure parameters as regulatory factors; Output layer: output pixel coordinate position (x, y), or spatial disturbance component Δx, Δy.

[0041] The known arrangement area is used for supervised learning in the training process, and the loss function is the Euclidean distance error between the output coordinates and the real blueprint coordinates. To enhance the generalization ability, a mixed data training strategy can be used, such as introducing a small amount of artificial interference (such as temperature gradient change, local shadow) to enhance the stability of the sample. The final model will have the ability to "from thermal response dynamic behavior → predict spatial position", that is, it can automatically infer the accurate position of the pixel point in the actual space from the response atlas, overcoming the influence of factors such as microscope distortion and image displacement. After the neural modeling is completed, the modeling network can be applied to each pixel point in the abnormal response area, and the thermal disturbance response curve L(t) of the pixel point is input into the model to obtain its predicted coordinates (x̂, ) in the physical space. This method is more robust than simply based on image coordinate inversion, and is especially suitable for positioning tasks in the presence of small position drift, non-uniform thermal diffusion, local imaging distortion, etc. The predicted coordinate results of all abnormal pixel points form a set of physical space coordinate sets S_abn = {(x̂1, 1), (x̂2, 2),... (x̂n, n)}, which is the final positioning of the abnormal pixel array coordinate graph. According to actual needs, it can be converted into pixel number, panel chip number, array number, etc. to complete the traceability of the abnormal result.

[0042] At the same time, the following detection report output data can be generated: Pixel coordinate range of the abnormal area (minimum rectangular bounding box); Abnormal pixel point number and density; Offset mean / max from standard blueprint; Abnormal area center position and standard pixel layout comparison graph.

[0043] In this embodiment, step S5 includes the following steps: Obtain a preset pixel point design arrangement blueprint, and calculate pixel transfer error for the actual pixel point coordinates to extract position offset and direction angle deviation; Perform pixel color channel recognition on the abnormal deviation display module area, and perform material response difference analysis to obtain material response difference characteristics of different color channels; Based on the material response difference characteristics, transfer error attribution analysis is performed on the light response difference tensor matrix to obtain transfer error factors of different channels; According to the transfer error factors of different channels, self-adaptive pixel transfer error compensation is performed on the position offset and direction angle deviation, thereby generating transfer error compensation parameters.

[0044] In this embodiment, the Micro LED micro display module has a fixed arrangement order and coordinate position for each pixel point during design. These information is usually provided by the process design team in the form of arrangement blueprint, which records the accurate position, spacing and color arrangement rule of the pixels in the module. Based on the above steps, the response coordinates of each pixel point in the actual module under thermal excitation are obtained. Now, these response points need to be paired and compared with the design blueprint. The conversion and correction between coordinate systems are performed to ensure that the coordinates in the detection image and the design blueprint are in the same scale and direction. For each actual detected pixel point and its theoretical design position, the spatial offset between them is extracted to form the position offset data. At the same time, the offset angle of each pixel is calculated by combining the direction of the displacement vector, which clearly shows the offset trend during the packaging process. The output results of this step are the spatial offset distance and offset direction angle of each abnormal pixel point, which are used as input for the subsequent compensation model. According to the design logic of the Micro LED module, different color channels use different types of chip materials and packaging processes, for example, the red channel usually uses AlInGaP material, and the green and blue channels use InGaN material with different wavelengths. Therefore, there are physical differences in thermal conductivity, packaging stress response, current driving characteristics, etc. among different channels. The abnormal offset area in the detection result is classified by color channel, and the RGB channel attribution of each pixel is identified according to the preset arrangement rule or spectral response characteristics under thermal excitation. Then, the response parameters of each channel pixel during thermal disturbance are extracted, such as response delay time, light emission start time, light intensity rise rate, maximum light intensity, response stability, etc. By comparing and analyzing, the significant differences in material response of different channels can be identified, such as the response start lag, poor light emission stability, and large thermal coupling fluctuation of a certain color channel under the same thermal excitation conditions. These features constitute the technical characterization of material response difference, which is an important input for the error attribution model. Based on the identification of the thermal response feature differences of different channels, the differences should be analyzed corresponding to the actual offset behavior of the pixels to determine the key physical causes leading to the transfer error. For the abnormal pixels in each color channel, a data association structure is constructed, which includes response features (such as response delay, curvature fluctuation, light intensity change trend, etc.) and spatial offset behavior (position offset, direction angle). Through statistical analysis, cluster analysis or regression modeling, etc., the correlation between response features and position offset is evaluated, and the indicators with significant causal relationship are identified. If the response delay of the blue channel is significantly higher than that of other channels, and the corresponding offset amount shows a centralized trend, it can be preliminarily determined that the delay response is the main cause of the offset of the blue channel. Further combined with the process conditions and material properties, the packaging stress release mechanism or thermal expansion inconsistency of this channel can be determined as the main error source.This stage outputs a "List of Transfer Error Influence Factors" for each color channel, clearly indicating the main material or process characteristics that cause spatial offset in each channel, and establishing a physical attribution model for the offset error. Based on the attribution model, an adaptive compensation mechanism applicable to different color channels and response states needs to be established to correct errors in the pixel positions of abnormal regions. The compensation mechanism uses color channels, response behavior, and positional offset characteristics as input variables to construct error inverse correction rules for specific channels. Different compensation strategies are used for different channels, for example: For the blue channel dominated by delayed response, linear displacement compensation in the same direction is implemented; For the red channel dominated by response drift, a trend compensation method based on local mean fitting is adopted; For the green channel dominated by stability anomalies, fluctuation constraints are introduced for dynamic correction.

[0045] For each anomalous pixel, a corresponding compensation vector is calculated based on its response characteristics and channel affiliation, including position compensation and angle correction values. This compensation vector is then combined with the actual detected coordinates to generate a corrected position estimation result. All compensation parameters are output as a structured data table, containing fields such as pixel index, original coordinates, corrected coordinates, channel information, compensation factor, and error confidence level. This parameter table can be used for: Pixel error visualization analysis; Process repair guidance; Automatic alignment system dynamic calibration; Defect tracking and tracing database input.

[0046] In this embodiment, step S6 includes the following steps: Pixel distribution is fitted based on transfer error compensation parameters to construct a transfer error compensation distribution map. Based on the transfer error compensation distribution map, the mass transfer process parameters are optimized and simulated to generate multiple mass transfer process simulation parameters; the simulation parameters include adsorption pressure, release acceleration, temperature curve and calibration time sequence. Mass transfer accuracy is evaluated for multiple mass transfer process simulation parameters, and the optimal mass transfer process simulation parameters are extracted. We employ deep reinforcement learning to perform dynamic packaging control on the optimal mass transfer process simulation parameters, and construct an intelligent packaging detection optimization model.

[0047] In this embodiment, the compensated coordinates of each pixel are constructed based on the previously acquired transfer error compensation parameter data (such as position offset, direction angle deviation, etc.). The compensated pixel position represents the final position of each pixel during the thermal disturbance or encapsulation process.

[0048] Fitting method selection: Spatial distribution analysis is performed on all compensated pixel coordinate data. By using fitting techniques such as regression analysis, cluster analysis, or least squares method, the pixel point distribution of the entire module is modeled. Different fitting methods can be selected, such as linear regression, curve fitting, or high-dimensional space interpolation, and the most appropriate fitting model is selected according to the complexity of the pixel distribution.

[0049] Atlas generation: During the fitting process, factors such as the spatial position of each pixel point, position offset, compensation value, etc. are considered and spatial interpolation, smoothing processing and data visualization are performed. The final "transfer error compensation distribution atlas" shows the transfer error compensation of each pixel point in the entire module, providing a basis for subsequent process optimization. According to the information in the transfer error compensation distribution atlas, process parameters closely related to transfer accuracy are selected. These key parameters include: adsorption pressure: this parameter defines the adsorption force between the pixel point and the substrate during the transfer process. Too low adsorption pressure may cause the pixel to be not firm, while too high adsorption pressure may cause pixel damage or uneven pressure distribution.

[0050] Release acceleration: The acceleration during the release process affects the release stability of the pixel. If the acceleration is too large, it may cause pixel offset; if it is too small, it may cause slow transfer speed.

[0051] Temperature curve: Temperature change has an important influence on the thermal expansion and stress release of materials, and a reasonable temperature curve can ensure the stability and accuracy of the pixel during the transfer process.

[0052] Calibration timing: Calibration timing refers to the time control of each step of the process during the transfer process, ensuring that each step of the operation corresponds to the predetermined step, avoiding errors caused by time mismatch. By changing different process parameters, multiple sets of massive transfer process simulation parameters are constructed. The simulation process considers multiple variables such as adsorption pressure, release acceleration, temperature change, and calibration timing, ensuring that it can cover multiple possible process combinations. Numerical simulation tools (such as finite element analysis FEA) are used to simulate different process parameter combinations, simulating stress distribution, thermal stress coupling effect, mechanical response, etc. during the transfer process. Through these simulations, the performance of each set of process parameters in the transfer process is evaluated, including error, stability, efficiency, etc. Based on the simulation results, multiple sets of transfer process simulation parameters are generated. Each set of simulation parameters will correspond to a different process combination, covering the transfer effect under different adsorption pressures, release accelerations, temperature curves, etc. These simulation parameters provide data support for the next step of precision evaluation and optimization.

[0053] For the target of process simulation, set the precision evaluation standard. The main evaluation indicators include: Position accuracy: The accuracy of the pixel point transferred to the target position, whether there is a large offset.

[0054] Stability: repeatability and consistency during the transfer process, whether there is an error that changes over time or multiple operations.

[0055] Transfer efficiency: time and energy consumption required by the process, whether it can meet the production speed and energy saving requirements.

[0056] By analyzing the results of each set of process simulation parameters in detail, the key information such as transfer accuracy, error fluctuation, process time under each set of simulation parameters is extracted. By comparing the performance of different process combinations, the process parameters that can minimize errors, improve accuracy and stability are identified. According to the accuracy evaluation results, the best performing process simulation parameters are selected as the optimal mass transfer process simulation parameters. These optimal parameters will provide a reference for subsequent packaging process control and intelligent optimization. Based on the optimal process simulation parameters and transfer error compensation parameters, a deep reinforcement learning model is constructed. The task of this model is to automatically adjust process parameters such as adsorption pressure, release acceleration, temperature curve according to real-time feedback signals such as position error, thermal stress, mechanical response, etc.

[0057] The deep reinforcement learning model uses reinforcement learning algorithms such as Q-learning, deep Q-networks, etc. Through interaction with the environment, it continuously optimizes the control strategy, so that it can automatically adjust the operation parameters under dynamic changing process conditions, minimizing errors. Based on the control strategy trained by deep reinforcement learning, real-time dynamic control of the packaging process is realized. The system adjusts the process parameters in real time through sensors such as temperature, pressure, acceleration, displacement, etc. to ensure that each step is executed accurately and the transfer accuracy of the pixel points is optimized. Dynamic packaging control can automatically adjust process parameters according to environmental changes, material properties, equipment status, etc. to ensure high precision and stability of the transfer process. Through the control strategy optimized by deep reinforcement learning, an intelligent packaging detection optimization model is formed. This model can automatically evaluate the current process precision during production and detect and repair abnormal situations. This model can monitor every step of the packaging process in real time, automatically adjust process parameters to ensure accurate transfer of each pixel point, and ultimately achieve efficient and low-error packaging detection.

[0058] In this embodiment, a full-color Micro LED micro display module packaging detection device is provided for executing the method as described above, comprising: A thermal induction module for pixel-level thermal induction of the micro display module to be detected, collecting global pixel thermal induction spontaneous light response parameters, instantaneous light response analysis, and generating an instantaneous light response curvature. A response trend module for multi-time point response trend analysis based on the instantaneous light response curvature and global pixel thermal disturbance coupling evolution to construct a pixel point thermal disturbance response map. An abnormal area positioning module is configured to perform multi-dimensional response difference calculation between pixels based on a pixel point thermal disturbance response atlas and abnormal deviation area positioning, and extract an abnormal deviation display module area; A pixel point calculation module is configured to perform spatial arrangement inversion on the abnormal deviation display module area, perform accurate pixel point positioning calculation, and generate actual pixel point coordinates of the abnormal area; A transfer error compensation module is configured to obtain a preset pixel point design arrangement blueprint, perform pixel transfer error calculation and adaptive pixel transfer error compensation on the actual pixel point coordinates, and thus generate a transfer error compensation parameter; A transfer process optimization module is configured to perform massive transfer process parameter optimization simulation based on the transfer error compensation parameter, perform dynamic packaging control, and construct an intelligent packaging detection optimization model.

[0059] The application further provides a computer device comprising a memory and a processor, and the memory stores a computer program, and the processor implements the steps of the full-color MicroLED micro display module packaging detection method according to any one of the above embodiments when executing the computer program.

[0060] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, the scope of the application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the application file are intended to be included in the application.

[0061] The above description is only a specific embodiment of the application, enabling those skilled in the art to understand or implement the application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A full-color Micro LED micro-display module packaging detection method, characterized in that, The method comprises the following steps: Step S1: pixel-level thermal induction is performed on the micro display module to be detected, global pixel thermal induction self-luminous response parameters are collected, transient light response analysis is performed, and transient light response curvature is generated; Step S2: multi-time point response trend analysis is performed based on the transient light response curvature, and global pixel thermal disturbance coupling evolution is performed to construct a pixel point thermal disturbance response map; Step S3: multi-dimensional response difference calculation and abnormal deviation region positioning between pixels are performed according to the pixel point thermal disturbance response map, and an abnormal deviation display module region is extracted; Step S4: spatial arrangement inversion is performed on the abnormal deviation display module region, and accurate pixel point positioning calculation is performed to generate actual pixel point coordinates of the abnormal region; Step S5: a preset pixel point design arrangement blueprint is obtained, pixel transfer error calculation and adaptive pixel transfer error compensation are performed on the actual pixel point coordinates, and a transfer error compensation parameter is generated; Step S6: based on the transfer error compensation parameter, a large amount of transfer process parameter optimization simulation is performed, dynamic packaging control is performed, and an intelligent packaging detection optimization model is constructed. 2.The full-color Micro LED micro-display module package detection method according to claim 1, wherein, The specific steps of step S1 are as follows: Different thermal field excitation signals are defined, pixel-level thermal induction is performed on the micro display module to be detected, and global pixel thermal induction self-luminous response parameters are collected; Based on the self-luminous response parameters, initial light emission intensity calculation is performed on each RGB sub-pixel point, and the initial light emission intensity of the pixel point under different thermal fields is extracted; According to the initial light emission intensity of the pixel point, time sequence light intensity change tracking is performed to obtain the dynamic light intensity change curve of each pixel point; Based on the pixel-level thermal induction self-luminous response log, a thermal field loading first timestamp and an initial light emission timestamp are calculated; According to the thermal field loading first timestamp and the initial light emission timestamp, the thermal excitation response delay time of each sub-pixel point is calculated; According to the thermal excitation response delay time, transient light response analysis under thermal field disturbance is performed on the dynamic light intensity change curve to generate transient light response curvature. 3.The full-color Micro LED micro-display module package detection method according to claim 1, wherein, The specific steps of step S2 are as follows: Based on the transient light response curvature, the positive and negative response values of each pixel point are identified; Multi-time point response trend analysis is performed on the positive and negative response values to obtain the thermal disturbance response characteristics of each pixel point, the thermal disturbance response characteristics including advanced type, lagging type and abnormal type; Based on the thermal disturbance response characteristics, response type definition is performed to obtain the response type label of each pixel point; Full time sequence light intensity trend mining is performed on the dynamic light intensity change curve to obtain the full time sequence light intensity trend characteristics of each pixel point; Based on the response type label, each pixel point is labeled, and display module global pixel thermal disturbance coupling evolution is performed according to the full time sequence light intensity trend characteristics to construct a pixel point thermal disturbance response map.

4. The full-color Micro LED micro-display module packaging detection method according to claim 1, characterized in that, The specific steps of step S3 are as follows: Identify the same color channel pixel points of the micro display module to be detected; According to the pixel point thermal disturbance response map, adjacent pixel division is performed on the same color channel pixel points, and multi-dimensional response difference calculation between pixels is performed to obtain a light response difference tensor matrix between adjacent pixels; Multi-dimensional response deviation calculation is performed on the light response difference tensor matrix to extract multi-dimensional response deviation values; According to the preset response difference distance threshold, the multi-dimensional response deviation value is subjected to abnormal response identification, and abnormal deviation area positioning is performed, and an abnormal deviation display module area is extracted.

5. The full-color Micro LED micro-display module packaging detection method according to claim 1, characterized in that, The specific steps of step S4 are: extracting a light response difference tensor matrix of the abnormal deviation display module area; According to the preset pixel point design arrangement blueprint, the light response difference tensor matrix is subjected to space arrangement inversion to obtain a current pixel point arrangement diagram; Physical perception neural modeling is performed on the pixel point thermal disturbance response spectrum to obtain a mapping rule of thermal disturbance response and spatial coordinates; Based on the mapping rule, accurate pixel point positioning calculation is performed on the current pixel point arrangement diagram to generate actual pixel point coordinates of the abnormal area. 6.The full-color Micro LED micro-display module package detection method according to claim 1, wherein, The specific steps of step S5 are: obtain a preset pixel point design arrangement blueprint, and perform pixel transfer error calculation on the actual pixel point coordinates to extract a position offset and a direction angle deviation; perform pixel color channel identification on the abnormal deviation display module area, and perform material response difference analysis to obtain material response difference characteristics of different color channels; Based on the material response difference characteristics, transfer error attribution analysis is performed on the light response difference tensor matrix to obtain transfer error factors of different channels; According to the transfer error factors of different channels, self-adaptive pixel transfer error compensation is performed on the position offset and the direction angle deviation, thereby generating a transfer error compensation parameter.

7. The full-color Micro LED micro-display module packaging detection method according to claim 1, wherein The specific steps of step S6 are: Based on the transfer error compensation parameter, pixel point distribution fitting is performed to construct a transfer error compensation distribution map; According to the transfer error compensation distribution map, a large amount of transfer process parameter optimization simulation is performed to generate a plurality of large amount of transfer process simulation parameters; the simulation parameters include adsorption pressure, release acceleration, temperature curve and calibration timing; A large amount of transfer precision evaluation is performed on the plurality of large amount of transfer process simulation parameters to extract an optimal large amount of transfer process simulation parameter; Deep reinforcement learning is performed on the optimal large amount of transfer process simulation parameter, and dynamic packaging control is performed to construct an intelligent packaging detection optimization model. 8.A full-color Micro LED micro-display module packaging detection device, characterized in that, A full-color Micro LED micro display module packaging detection method is used to perform the method of claim 1, comprising: a thermal induction module for pixel-level thermal induction of a micro display module to be detected, collecting global pixel thermal induction spontaneous light emission response parameters, instantaneous light response analysis, and generating an instantaneous light response curvature; a response trend module for multi-time point response trend analysis based on the instantaneous light response curvature, and global pixel thermal disturbance coupling evolution to construct a pixel point thermal disturbance response spectrum; an abnormal area positioning module for pixel-to-pixel multi-dimensional response difference calculation and abnormal deviation area positioning based on the pixel point thermal disturbance response spectrum to extract an abnormal deviation display module area; a pixel point calculation module for space arrangement inversion of the abnormal deviation display module area and accurate pixel point positioning calculation to generate actual pixel point coordinates of the abnormal area; a transfer error compensation module for obtaining a preset pixel point design arrangement blueprint and performing pixel transfer error calculation and self-adaptive pixel transfer error compensation on the actual pixel point coordinates to generate a transfer error compensation parameter; A transfer process optimization module is used for performing mass transfer process parameter optimization simulation based on transfer error compensation parameters and dynamic packaging control, and constructing an intelligent packaging detection optimization model. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8. The processor implements the steps of the full-color Micro LED micro display module packaging detection method of any one of claims 1 to 7 when executing the computer program.

Images