A method and system for detecting a micro-led chip array

Abstract

Embodiments of the present application provide a MicroLED chip array detection method and system, the method comprising: acquiring a display image of a MicroLED chip array to be detected; preprocessing the display image to obtain a standardized image matrix; performing region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region; based on the structured data, extracting a multi-dimensional feature vector from each chip region; inputting the multi-dimensional feature vector of each chip region into a pre-trained deep neural network model to obtain a corresponding reconstructed feature vector; calculating the reconstruction error between the multi-dimensional feature vector and the reconstructed feature vector of each chip region; and marking a chip region with a reconstruction error greater than a preset threshold as an abnormal region. Through the method of the embodiments of the present application, abnormal detection and rapid positioning of the MicroLED chip array can be achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of detection, in particular to a detection method and system of a MicroLED chip array. BACKGROUND

[0002] With the rapid development of MicroLED (Micro Light Emitting Diode) display technology, its advantages in high brightness, low power consumption, high contrast and long service life make it gradually become an important direction of the next generation of display devices. However, the size of the MicroLED chip is usually in the order of 10-100 microns, and the photoelectric performance and structural characteristics of a single light emitting unit are easily affected by micro-defects in the process, such as grain cracks, foreign matter pollution, uneven brightness, open circuit or short circuit defects. Therefore, how to quickly detect defects of the MicroLED chip has become a problem to be solved. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide a detection method and system of a MicroLED chip array, to solve the problem of how to quickly detect defects of the MicroLED chip. The specific technical solutions are as follows:

[0004] In a first aspect of the embodiments of the present application, a optical detection method of a MicroLED chip array is provided, comprising:

[0005] obtaining a display image of a MicroLED chip array to be detected;

[0006] preprocessing the display image to obtain a standardized image matrix;

[0007] performing region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region, the structured data at least including center coordinates and contour boundary information;

[0008] extracting a multi-dimensional feature vector from each chip region based on the structured data, the multi-dimensional feature vector including a local gray histogram, a texture feature and a morphological feature;

[0009] inputting the multi-dimensional feature vector of each chip region into a pre-trained deep neural network model to obtain a corresponding reconstructed feature vector;

[0010] calculating a reconstruction error between the multi-dimensional feature vector and the reconstructed feature vector of each chip region; and marking a chip region with a reconstruction error greater than a preset threshold as an abnormal region.

[0011] In one possible implementation, the preprocessing of the displayed image to obtain a normalized image matrix includes:

[0012] The displayed image is sequentially processed by brightness equalization, gamma correction, noise suppression, and image normalization.

[0013] The pixel values ​​are mapped to the [0,1] interval to generate the standardized image matrix.

[0014] In one possible implementation, the step of performing region segmentation processing on the standardized image matrix to extract structured data for each MicroLED chip region includes:

[0015] The standardized image matrix is ​​binarized to obtain a mask matrix;

[0016] Based on the contour extraction algorithm, connected component analysis is performed on the mask matrix to identify the MicroLED chip region;

[0017] The center point coordinates and contour boundary information of each identified chip region are calculated to form the structured data;

[0018] The structured data includes the center point coordinates, contour boundary point sequence, and region identifier for each chip region, and is represented as Region={Center(xc, yc),Contour[Pi(x, y)], Index(ID)}, where Center(xc, yc) are the chip center point coordinates; Contour[Pi(x, y)] is the chip contour boundary point sequence; and Index(ID) is the unique identifier for the chip region.

[0019] In one possible implementation, after calculating the reconstruction error between the multidimensional feature vector and the reconstructed feature vector for each chip region, and marking chip regions with reconstruction errors greater than a preset threshold as abnormal regions, the method further includes:

[0020] Obtain the bounding box coordinates corresponding to the abnormal region;

[0021] Non-maximum suppression is performed on the bounding box coordinates to remove overlapping false detection boxes;

[0022] Morphological filtering is applied to the results after nonmaximum suppression to further eliminate false detections caused by noise;

[0023] A defect mask layer is generated based on the processed bounding box coordinates.

[0024] In one possible implementation, after generating the defect mask layer based on the processed bounding box coordinates, the method further includes:

[0025] The defect mask layer and the display image are superimposed to obtain a detection result image with defect markings;

[0026] The structured result data, which includes defect area identification, confidence level, and location information, is stored in the database.

[0027] A second aspect of this application provides a detection system for a MicroLED chip array, the system comprising:

[0028] An optical imaging module is used to acquire display images of the MicroLED chip array to be tested;

[0029] The image processing and analysis module is used to preprocess the displayed image to obtain a standardized image matrix; perform region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region, wherein the structured data includes at least center coordinates and contour boundary information; and extract multidimensional feature vectors from each chip region based on the structured data, wherein the multidimensional feature vectors include local grayscale histograms, texture features, and morphological features.

[0030] The machine learning detection module is used to input the multidimensional feature vectors of each chip region into a pre-trained deep neural network model to obtain the corresponding reconstructed feature vectors; calculate the reconstruction error between the multidimensional feature vectors and the reconstructed feature vectors of each chip region; and mark chip regions with reconstruction errors greater than a preset threshold as abnormal regions.

[0031] In one possible implementation, the system further includes:

[0032] The motion control module is used to drive the MicroLED chip array to be tested to move in three dimensions in order to perform point-by-point scanning of the MicroLED chip array to be tested.

[0033] In one possible implementation, the image processing and analysis module is specifically used to sequentially perform brightness equalization, gamma correction, noise suppression, and image normalization on the displayed image; map pixel values ​​to the [0,1] interval, and generate the normalized image matrix.

[0034] In one possible implementation, the image processing and analysis module is specifically used to binarize the standardized image matrix to obtain a mask matrix; perform connected component analysis on the mask matrix based on a contour extraction algorithm to identify the MicroLED chip region; calculate the center point coordinates and contour boundary information of each identified chip region to form the structured data; the structured data includes the center point coordinates, contour boundary point sequence, and region identifier of each chip region, and the structured data is represented as Region={Center(xc, yc),Contour[Pi(x,y)], Index(ID)}, where Center(xc, yc) is the chip center point coordinates; Contour[Pi(x, y)] is the chip contour boundary point sequence; and Index(ID) is the unique identifier of the chip region.

[0035] In one possible implementation, the system further includes:

[0036] The result visualization and data storage module is used to obtain the bounding box coordinates corresponding to the abnormal region; perform non-maximum suppression processing on the bounding box coordinates to remove overlapping false detection boxes; perform morphological filtering on the results after non-maximum suppression processing to further eliminate false detections caused by noise; and generate a defect mask layer based on the processed bounding box coordinates.

[0037] In one possible implementation, the result visualization and data storage module is further configured to overlay the defect mask layer and the display image to obtain a detection result image with defect markings; and to store the structured result data containing defect region identification, confidence level and location information into a database.

[0038] Another aspect of the embodiments of this application also provides an electronic device, including:

[0039] Memory, used to store computer programs;

[0040] The processor, when executing the program stored in the memory, performs the detection of any of the aforementioned MicroLED chip arrays.

[0041] In another aspect of the embodiments of this application, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the detection of any of the MicroLED chip arrays described above.

[0042] In another aspect of the embodiments of this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform the detection of any of the MicroLED chip arrays described above.

[0043] Beneficial effects of the embodiments in this application:

[0044] This application provides a method and system for detecting MicroLED chip arrays. The method includes: acquiring a display image of the MicroLED chip array to be detected; preprocessing the display image to obtain a standardized image matrix; performing region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region, wherein the structured data includes at least center coordinates and contour boundary information; extracting a multidimensional feature vector from each chip region based on the structured data, wherein the multidimensional feature vector includes a local grayscale histogram, texture features, and morphological features; inputting the multidimensional feature vector of each chip region into a pre-trained deep neural network model to obtain a corresponding reconstructed feature vector; calculating the reconstruction error between the multidimensional feature vector and the reconstructed feature vector of each chip region; and marking chip regions with reconstruction errors greater than a preset threshold as abnormal regions. The method of this application embodiment can, after acquiring the display image of the MicroLED chip array to be detected, perform region segmentation processing on the standardized image matrix to extract the structured data of each MicroLED chip region, and then use a pre-trained neural network model to calculate the reconstruction error between the multidimensional feature vector and the reconstructed feature vector of each chip region, thereby determining abnormal regions and realizing the anomaly detection and rapid localization of the MicroLED chip array.

[0045] Of course, implementing any product or method of this application does not necessarily require achieving all of the advantages described above at the same time. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other embodiments can be obtained based on these drawings.

[0047] Figure 1 This is a schematic flowchart of a detection method for a MicroLED chip array provided in an embodiment of this application;

[0048] Figure 2 This is another schematic flowchart illustrating the detection method for MicroLED chip arrays provided in the embodiments of this application;

[0049] Figure 3 A schematic diagram of a detection system for a MicroLED chip array provided in an embodiment of this application;

[0050] Figure 4 Another schematic diagram of the detection system for the MicroLED chip array provided in the embodiments of this application;

[0051] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art based on this application are within the scope of protection of this application.

[0053] The first aspect of this application provides a method for detecting MicroLED chip arrays, see [link to relevant documentation]. Figure 1 , Figure 1 This is a schematic flowchart of a detection method for a MicroLED chip array provided in an embodiment of this application. The method includes:

[0054] Step S11: Obtain the display image of the MicroLED chip array to be tested;

[0055] Step S12: Preprocess the displayed image to obtain a standardized image matrix;

[0056] Step S13: Perform region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region. The structured data includes at least center coordinates and contour boundary information.

[0057] Step S14: Based on the structured data, extract multi-dimensional feature vectors from each chip region. The multi-dimensional feature vectors include local grayscale histograms, texture features, and morphological features.

[0058] Step S15: Input the multidimensional feature vectors of each chip region into the pre-trained deep neural network model to obtain the corresponding reconstructed feature vectors;

[0059] Step S16: Calculate the reconstruction error between the multidimensional feature vector and the reconstructed feature vector for each chip region; mark chip regions with reconstruction errors greater than a preset threshold as abnormal regions.

[0060] Corresponding to step S11 above, when acquiring the display image of the MicroLED chip array to be tested, a high-resolution optical image of the MicroLED chip array can be obtained through an industrial camera. The acquired analog light signal is converted into a digital signal stream via an image acquisition card and cached in the image data area of ​​memory. Specifically, high-resolution non-contact imaging of the MicroLED chip can be performed using an industrial camera, a telecentric lens, a ring light source, and a light source controller. The telecentric lens ensures stable magnification and low distortion. Optimal contrast can be achieved for materials with different reflective properties by adjusting the light intensity. In practical use, the sample can be driven to move precisely in the XYZ three-dimensional direction to achieve point-by-point scanning detection of the chip array. Synchronization with the camera via trigger signals can be controlled to ensure consistent image acquisition time at each detection position.

[0061] Corresponding to step S12 above, the displayed image is preprocessed to obtain a standardized image matrix. Specifically, during preprocessing, operations such as grayscale equalization, noise filtering, gamma correction, and region of interest extraction can be performed to provide input for the subsequent machine learning module.

[0062] Corresponding to step S13 above, when performing region segmentation processing on the standardized image matrix and extracting structured data for each MicroLED chip region, the structured data includes at least center coordinates and contour boundary information.

[0063] Corresponding to step S14 above, when extracting multi-dimensional feature vectors from each chip region based on the structured data, local gray-level histograms, texture features, and morphological features can be extracted to obtain feature vectors for each chip. The multi-dimensional feature vectors include local gray-level histograms, texture features, and morphological features.

[0064] Corresponding to step S15 above, the multidimensional feature vectors of each chip region are input into a pre-trained deep neural network model to obtain the corresponding reconstructed feature vectors. Specifically, the feature vectors F of each chip can be input into a pre-trained deep neural network model, and forward propagation can be performed through the model to generate reconstructed features F′.

[0065] Corresponding to step S16 above, the reconstruction error between the multidimensional feature vector and the reconstructed feature vector for each chip region is calculated; chip regions with reconstruction errors greater than a preset threshold are marked as abnormal regions. The reconstruction error between the multidimensional feature vector and the reconstructed feature vector is calculated as E = |F - F′|; if the error E > τ (the set threshold), the corresponding region is marked as an abnormal region. In actual use, the output abnormal information may include: Region_ID (identifier corresponding to the abnormal region), Defect_Score (abnormal probability), and Bounding_Box (bounding box). In one example, the model output data structure is: Result = {Region_ID, Defect_Score, Bounding_Box}.

[0066] As can be seen, the method of this application embodiment can, after acquiring the display image of the MicroLED chip array to be detected, perform region segmentation processing on the standardized image matrix to extract the structured data of each MicroLED chip region, and then use a pre-trained neural network model to calculate the reconstruction error between the multidimensional feature vector and the reconstructed feature vector of each chip region, thereby determining the abnormal region and realizing the anomaly detection and rapid localization of the MicroLED chip array.

[0067] In one possible implementation, the preprocessing of the displayed image to obtain a standardized image matrix includes: sequentially performing brightness equalization, gamma correction, noise suppression, and image normalization on the displayed image; mapping pixel values ​​to the [0,1] interval to generate the standardized image matrix. Specifically, the image undergoes standardized preprocessing, including: brightness equalization, gamma correction, noise suppression (using adaptive bilateral filtering), and image normalization, mapping pixel values ​​to [0,1]. The preprocessed image is stored as a matrix I′(x, y, c), which serves as input to the subsequent feature extraction module. Brightness equalization can redistribute pixel brightness values ​​to make the image's light and dark distribution more uniform; for example, darker areas are brightened, and overly bright areas are dimmed, thereby improving overall contrast and detail visibility. Gamma correction is an inverse operation method in image systems that performs nonlinear calculations on luminance or tri-color stimulus values, mainly used to adjust the nonlinear hue of an image. Adaptive bilateral filtering for noise suppression can achieve edge-preserving denoising by comprehensively considering spatial proximity (Gaussian weight) and pixel value similarity (range weight). Image normalization refers to the process of performing a series of standard processing transformations on an image to transform it into a fixed standard form. This standard image is called a normalized image.

[0068] In one possible implementation, the step of performing region segmentation processing on the standardized image matrix to extract structured data for each MicroLED chip region includes: binarizing the standardized image matrix to obtain a mask matrix; performing connected component analysis on the mask matrix based on a contour extraction algorithm to identify the MicroLED chip regions; calculating the center point coordinates and contour boundary information of each identified chip region to form the structured data; the structured data includes the center point coordinates, contour boundary point sequence, and region identifier for each chip region, and the structured data is represented as Region={Center(xc, yc),Contour[Pi(x, y)], Index(ID)}, where Center(xc, yc) is the chip center point coordinates; Contour[Pi(x, y)] is the chip contour boundary point sequence; and Index(ID) is the unique identifier for the chip region. Specifically, a threshold segmentation algorithm can be used to binarize I′(x, y, c) to generate a mask matrix M(x, y). Simultaneously, a contour extraction algorithm is used to identify the MicroLED chip regions, calculating the center point coordinates and contour boundaries of each chip to form structured data. Specifically, the center point coordinates of each chip can be represented by Center(xc, yc), the contour boundaries by Contour[Pi(x, y)], and the identification information of each chip by Index(ID). In one example, the structured data could be Region = {Center(xc, yc),Contour[Pi(x, y)], Index(ID)}, which can serve as the basic input for subsequent feature analysis modules.

[0069] Specifically, local gray-level histograms, texture features, and morphological features (area, aspect ratio) can be extracted from each chip region. A feature vector is generated for each region: F_i = [f_gray, f_texture, f_shape, f_position]. The feature matrix F = {F_1, F_2, ..., F_n} is stored in a feature buffer in memory. F_i is the feature vector generated for each region, where i is 1, 2, ..., n; f_gray represents the local gray-level histogram; f_texture represents the texture feature; f_shape represents the morphological feature; f_position represents the positional feature; and F represents the matrix composed of the feature vectors of each chip. The local gray-level histogram, as a function of gray-level distribution, can statistically analyze the gray-level distribution in the image. Image texture features can extract the global characteristics of the object's surface structure by analyzing the gray-level or color distribution patterns of pixels in the image. Morphological features in an image are a set of attributes used in image processing to describe the shape, texture, and spatial distribution of an object, and can include four dimensions: color, texture, shape, and spatial relationships.

[0070] In one possible implementation, after calculating the reconstruction error between the multidimensional feature vector and the reconstructed feature vector for each chip region, and marking chip regions with reconstruction errors greater than a preset threshold as abnormal regions, the method further includes: obtaining the bounding box coordinates corresponding to the abnormal regions; performing non-maximum suppression processing on the bounding box coordinates to remove overlapping false detection boxes; performing morphological filtering on the result after non-maximum suppression processing to further eliminate false detections caused by noise; and generating a defect mask layer based on the processed bounding box coordinates. Specifically, non-maximum suppression and morphological filtering can be performed on the abnormal regions in the detection results to remove noise false detections; a mask layer D(x, y) can be generated based on the bounding box coordinates, and defect marking boxes can be drawn on the original image.

[0071] In one possible implementation, after generating the defect mask layer based on the processed bounding box coordinates, the method further includes: overlaying the defect mask layer and the displayed image to obtain a detection result image with defect markers; and storing the structured result data containing defect region identifiers, confidence levels, and location information into a database. Specifically, the original image can be overlaid on D(x, y) to generate a detection result image with defect markers; simultaneously, the Result data can be stored in a database in a specified format to achieve traceable management of defect data. The final output includes: a detection result image; an anomaly statistics report; and a defect location information table.

[0072] In one example, see Figure 2 , Figure 2Another flowchart illustrating the detection method for a MicroLED chip array provided in this application embodiment includes: inputting an original MicroLED image; an image preprocessing module (threshold segmentation, image rotation correction); a contour extraction module extracting the outer contour and center point coordinates of the LED beads; an LED bead matrix generation module sorting rows and columns to construct a matrix relationship and outputting a matrix center coordinate table; a machine learning inference module calling the ONNX model (Open Neural Network Exchange, an open deep learning model standard format) to perform defect detection and outputting defect bounding boxes rectangles-offset; a region cutting module calculating each row and each column based on the matrix center coordinates to generate cutting parameters params; and a result output module outputting a data set including defect location and type.

[0073] A second aspect of this application provides a detection system for MicroLED chip arrays, see [link to relevant documentation]. Figure 3 , Figure 3 This is another schematic diagram of the detection system for a MicroLED chip array provided in an embodiment of this application. The system includes:

[0074] Optical imaging module 301 is used to acquire display images of the MicroLED chip array to be tested;

[0075] The image processing and analysis module 302 is used to preprocess the displayed image to obtain a standardized image matrix; perform region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region, wherein the structured data includes at least center coordinates and contour boundary information; and extract multidimensional feature vectors from each chip region based on the structured data, wherein the multidimensional feature vectors include local grayscale histograms, texture features, and morphological features.

[0076] The machine learning detection module 303 is used to input the multidimensional feature vectors of each chip region into a pre-trained deep neural network model to obtain the corresponding reconstructed feature vectors; calculate the reconstruction error between the multidimensional feature vectors and the reconstructed feature vectors of each chip region; and mark chip regions with reconstruction errors greater than a preset threshold as abnormal regions.

[0077] In one possible implementation, the system further includes:

[0078] The motion control module is used to drive the MicroLED chip array to be tested to move in three dimensions in order to perform point-by-point scanning of the MicroLED chip array to be tested.

[0079] In one possible implementation, the image processing and analysis module is specifically used to sequentially perform brightness equalization, gamma correction, noise suppression, and image normalization on the displayed image; map pixel values ​​to the [0,1] interval, and generate the normalized image matrix.

[0080] In one possible implementation, the image processing and analysis module is specifically used to binarize the standardized image matrix to obtain a mask matrix; perform connected component analysis on the mask matrix based on a contour extraction algorithm to identify the MicroLED chip region; calculate the center point coordinates and contour boundary information of each identified chip region to form the structured data; the structured data includes the center point coordinates, contour boundary point sequence, and region identifier of each chip region, and the structured data is represented as Region={Center(xc, yc),Contour[Pi(x,y)], Index(ID)}, where Center(xc, yc) is the chip center point coordinates; Contour[Pi(x, y)] is the chip contour boundary point sequence; and Index(ID) is the unique identifier of the chip region.

[0081] In one possible implementation, the system further includes:

[0082] The result visualization and data storage module is used to obtain the bounding box coordinates corresponding to the abnormal region; perform non-maximum suppression processing on the bounding box coordinates to remove overlapping false detection boxes; perform morphological filtering on the results after non-maximum suppression processing to further eliminate false detections caused by noise; and generate a defect mask layer based on the processed bounding box coordinates.

[0083] In one possible implementation, the result visualization and data storage module is further configured to overlay the defect mask layer and the display image to obtain a detection result image with defect markings; and to store the structured result data containing defect region identification, confidence level and location information into a database.

[0084] To illustrate the solutions of the embodiments of this application, the following description is provided in conjunction with a specific example. (See [link to example]). Figure 4 The overall system structure consists of an optical imaging module, a motion control module, an image processing and analysis module, a machine learning detection module, and a result display and storage module. Each module is connected via an industrial Ethernet or high-speed data bus to achieve closed-loop control of signal acquisition, data transmission, and detection result output.

[0085] The optical imaging module includes an industrial camera, a telecentric lens, a ring light source, and a light source controller. This module is used for high-resolution, non-contact imaging of MicroLED chips. The telecentric lens ensures stable magnification and minimal distortion. The light source controller provides adjustable illumination intensity to achieve optimal contrast for materials with different reflective properties.

[0086] The motion control module, including a high-precision electric displacement platform and a motion control card, is used to drive the sample to move precisely in the XYZ three-dimensional directions, enabling point-by-point scanning detection of the chip array. The control card and camera are synchronized via trigger signals to ensure consistent image acquisition time at each detection position.

[0087] The image processing and analysis module performs image preprocessing procedures, including grayscale equalization, noise filtering, gamma correction, and region of interest extraction. The output of this module is a standardized image matrix, which serves as input for subsequent machine learning modules.

[0088] The machine learning detection module, the core of the system, consists of deep neural network algorithms (such as convolutional autoencoders, self-supervised contrastive learning models, or lightweight network models). Its main functions are: extracting features from the input standardized image; calculating reconstruction error or feature bias; determining anomalies based on thresholds or probability distribution models (such as Gaussian mixture models); and outputting defect type, confidence level, and spatial coordinates.

[0089] The results display and storage module includes a human-computer interaction interface, a database storage unit, and visualization tools. This module displays the inspection results in real time, providing defect location maps, statistical analysis reports, and production batch traceability functions. Inspection data is stored in the database in a structured format.

[0090] When the system runs, the execution process is as follows:

[0091] Step S1: The motion control module drives the detection platform to move to the designated detection area and sends a trigger signal to the optical imaging module.

[0092] Step S2: The optical imaging module acquires an image of the MicroLED chip and sends it to the processing unit via a high-speed interface (GigE).

[0093] Step S3: The image processing module executes a preprocessing algorithm to generate a standardized image matrix.

[0094] Step S4: The machine learning detection module calls the trained model to infer the input image and obtain the defect type and location.

[0095] Step S5: The results display module will visualize the detection information and save it to the database.

[0096] Step S6: If the detection result exceeds the set threshold, the system automatically sends an alarm signal to the production control system to achieve closed-loop feedback.

[0097] This system achieves non-contact, high-speed, and high-precision automatic detection of surface defects in MicroLED chips (such as dead pixels, abnormal brightness, contamination, and cracks). Specific hardware implementation effects include: improved image acquisition stability, avoiding contamination caused by manual contact; accelerated detection time per frame reduced to less than 100ms through experimental inference; synchronous triggering of the motion control system ensuring spatial positioning error is less than 5μm; and automatic screening and data traceability on the production line, reducing labor costs by approximately 80%. Traditional MicroLED detection typically relies on probe contact, electrical performance testing, or manual microscopic observation, resulting in low detection efficiency and easy sample damage. This invention acquires the reflected image of the sample surface through a non-contact optical imaging module, combined with a telecentric lens and multi-angle ring light source, achieving uniform illumination and high-contrast imaging, effectively avoiding damage and deviation caused by mechanical contact, fundamentally improving the safety and repeatability of the detection. This invention employs a deep learning feature extraction model to perform feature space modeling of MicroLED pixel-level defects. Compared to traditional threshold- or template-matching algorithms, this model can adapt to different lighting conditions, chip arrangement deviations, and background noise. Through feature learning, it achieves highly robust anomaly detection capabilities, increasing detection accuracy from approximately 85% to over 98%. This invention enables batch parallel image processing during the inference stage. Compared to traditional serial processor computation, the detection speed is increased by approximately 3-5 times, completing single-frame image detection within 100ms, meeting the real-time detection requirements of production lines. Self-supervised learning generates pseudo-defect samples, enabling model training based solely on normal samples. Simultaneously, a dynamic illumination equalization algorithm ensures stable detection performance under varying brightness and background conditions. This technical feature, combined with machine learning algorithms, gives the system continuous self-learning and transfer adaptation capabilities, maintaining consistent performance across production lines or different equipment. The system architecture of this invention adopts a modular design, including an optical acquisition module, an image processing module, a machine learning detection module, a result visualization module, and a closed-loop control module. The modules interact through standardized data interfaces, ensuring good maintainability and scalability, facilitating integration into industrial control systems. This invention introduces a structured data storage and statistical analysis mechanism at the detection output end, which can automatically generate defect distribution heatmaps and process trend reports; and records the detection history of each chip through a database, realizing a traceable quality management closed loop and providing data support for yield optimization in the production process. In terms of hardware selection, this invention uses lightweight industrial cameras and embedded devices, significantly reducing system power consumption and cost while ensuring performance; compared with traditional high-performance server solutions, energy consumption is reduced by approximately 40%, and the overall system cost is reduced by approximately 30%. Experimental verification on a MicroLED packaging production line shows that the detection system of this invention can detect thousands of MicroLED chips per minute with a single device, achieving micron-level detection accuracy and a yield consistency exceeding 99%.The system has good environmental adaptability and can operate stably in industrial environments with large vibrations and temperature and humidity variations.

[0098] As can be seen, the system of this application embodiment can obtain the feature vector of the chip after acquiring the display image of the MicroLED chip array to be detected, through feature extraction and structured information extraction, and then use a pre-trained neural network model to identify and calculate the feature vector, thereby determining the abnormal area and realizing the abnormal detection and rapid localization of the MicroLED chip array.

[0099] This application also provides an electronic device, such as... Figure 5 As shown, it includes:

[0100] Memory 501 is used to store computer programs;

[0101] When processor 502 executes the program stored in memory 501, it performs the following steps:

[0102] Acquire a display image of the MicroLED chip array to be tested;

[0103] The feature matrix is ​​obtained by preprocessing the displayed image;

[0104] Based on the feature matrix, chip information is extracted to obtain structured data for each chip;

[0105] Based on the structured data of each chip, feature extraction is performed on each chip to obtain the feature vector of each chip;

[0106] The feature vectors of each chip are input into a pre-trained deep neural network model. The pre-trained deep neural network model calculates a reconstruction vector based on the feature vectors of each chip. For any region where a chip is located, the error between the feature vector and the reconstruction vector corresponding to that chip is calculated. When the error is greater than a preset threshold, the region corresponding to that chip is marked as an abnormal region.

[0107] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0108] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0109] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0110] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0111] In another embodiment provided in this application, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of the detection method for any of the MicroLED chip arrays described above.

[0112] In another embodiment provided in this application, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the detection methods for MicroLED chip arrays in the above embodiments.

[0113] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a solid-state drive (SSD), etc.

[0114] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0115] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system, electronic device, storage medium, and computer program product embodiments are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0116] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application are included within the scope of protection of this application.

Claims

1. A method for optical detection of a Micro LED chip array, characterized in that, The method comprises: acquiring a display image of a MicroLED chip array to be detected; preprocessing the display image to obtain a standardized image matrix; performing region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region, the structured data at least including center coordinates and contour boundary information; extracting a multi-dimensional feature vector from each chip region based on the structured data, the multi-dimensional feature vector including a local gray histogram, a texture feature, and a morphological feature; inputting the multi-dimensional feature vector of each chip region into a pre-trained deep neural network model to obtain a corresponding reconstructed feature vector; calling an ONNX model to calculate reconstruction errors between the multi-dimensional feature vector and the reconstructed feature vector of each chip region in parallel; and marking a chip region with a reconstruction error greater than a preset threshold as an abnormal region; the region segmentation processing on the standardized image matrix to extract structured data of each MicroLED chip region comprises: performing binaryzation processing on the standardized image matrix to obtain a mask matrix; performing connected domain analysis on the mask matrix based on a contour extraction algorithm to identify a MicroLED chip region; calculating center point coordinates and contour boundary information of each identified chip region to form the structured data; the structured data includes center point coordinates, contour boundary point sequences, and region identifiers of each chip region, and the structured data is expressed as Region={Center(xc, yc), Contour[Pi(x, y)], Index(ID)}, wherein Center(xc, yc) is the center point coordinates of the chip; Contour[Pi(x, y)] is the contour boundary point sequence of the chip; and Index(ID) is the unique identifier of the chip region.

2. The method of claim 1, wherein, the preprocessing of the display image to obtain a standardized image matrix comprises: sequentially performing brightness equalization, gamma correction, noise suppression, and image normalization processing on the display image; mapping pixel values to the [0, 1] interval to generate the standardized image matrix.

3. The method of claim 1, wherein, the calculation of the reconstruction errors between the multi-dimensional feature vector and the reconstructed feature vector of each chip region; after marking a chip region with a reconstruction error greater than a preset threshold as an abnormal region, the method further comprises: acquiring boundary box coordinates corresponding to the abnormal region; performing non-maximum suppression processing on the boundary box coordinates to remove overlapping false detection boxes; performing morphological filtering on the result after the non-maximum suppression processing to further eliminate false detections caused by noise; generating a defect mask layer based on the processed boundary box coordinates.

4. The method of claim 3, wherein, after generating the defect mask layer based on the processed boundary box coordinates, the method further comprises: superimposing the defect mask layer and the display image to obtain a detection result image with defect markers; storing structured result data including defect region identifiers, confidence levels, and position information into a database.

5. A detection system of a MicroLED chip array, characterized in that, the system comprises: An optical imaging module is configured to acquire a display image of a MicroLED chip array to be detected; An image processing and analysis module is configured to pre-process the display image to obtain a normalized image matrix, perform region segmentation processing on the normalized image matrix to extract structured data of each MicroLED chip region, and extract a multi-dimensional feature vector from each chip region based on the structured data, wherein the multi-dimensional feature vector includes a local grayscale histogram, a texture feature, and a morphological feature. A machine learning detection module is configured to input the multi-dimensional feature vector of each chip region into a pre-trained deep neural network model to obtain a corresponding reconstructed feature vector, call an ONNX model to calculate a reconstruction error between the multi-dimensional feature vector and the reconstructed feature vector of each chip region in parallel, and mark a chip region with a reconstruction error greater than a preset threshold as an abnormal region. The image processing and analysis module is specifically configured to perform binaryzation processing on the normalized image matrix to obtain a mask matrix, perform connected domain analysis on the mask matrix based on a contour extraction algorithm to identify a MicroLED chip region, calculate a center point coordinate and contour boundary information of each identified chip region to form the structured data, and represent the structured data as Region={Center(xc, yc), Contour[Pi(x, y)], Index(ID)}, wherein Center(xc, yc) is a chip center point coordinate, Contour[Pi(x, y)] is a chip contour boundary point sequence, and Index(ID) is a unique chip region identifier.

6. The system of claim 5, wherein, The system further comprises: A motion control module is configured to drive the MicroLED chip array to be detected to move in three-dimensional directions to perform point-by-point scanning on the MicroLED chip array to be detected.

7. The system of claim 5, wherein The image processing and analysis module is specifically configured to sequentially perform brightness equalization, gamma correction, noise suppression, and image normalization processing on the display image, map a pixel value to an interval [0, 1] to generate the normalized image matrix.

8. An electronic device, comprising: The system further comprises: A memory is configured to store a computer program; A processor is configured to execute the program stored on the memory to implement the method of any one of claims 1-4.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method of any one of claims 1-4. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method of any one of claims 1-4.

Images