A method and system for fault identification of LED displays

Abstract

This invention discloses a fault identification method and system for LED displays, relating to the field of display device status monitoring technology. The invention synchronously acquires high-frequency transient electrical signals, optical images, and distributed heat maps during the display's operation via a low-level hardware bus. It then uses affine transformation and resampling interpolation algorithms to map heterogeneous data to a unified physical pixel grid. The system constructs a digital twin theoretical benchmark based on real-time input source-end grayscale instructions. It extracts the photoelectric residual between measured and theoretical values ​​through differential operations, and combines photoelectric consistency verification and Pearson correlation analysis to eliminate video content interference. Simultaneously, it introduces transient temperature rise rate to adaptively compensate for the residual features using thermodynamic weights, and utilizes machine learning algorithms to match a hardware aging fingerprint database, ultimately outputting high-precision fault physical coordinates and failure mechanisms. This solution achieves non-intrusive intelligent operation and maintenance detection in dynamic playback scenarios.

Description

Technical Field

[0001] This invention belongs to the field of LED display screen fault detection, and particularly relates to an LED display screen fault identification method and system. Background Technology

[0002] The stable operation of LED displays is affected by multiple factors, including ambient temperature, power supply quality, and the aging of the light-emitting elements. Existing fault identification technologies are mainly divided into visual inspection based on external cameras and current readback detection based on driver chips. In conventional technical solutions, a camera is usually used to capture the display image and compare it with the original image, or a current threshold is set, triggering an alarm when the measured current deviates from the preset range.

[0003] However, conventional technologies have serious limitations in environments with complex interference: single optical detection cannot distinguish between environmental obstruction, surface contamination, and actual physical failure of the LED chips, and is susceptible to false alarms due to high refresh rate moiré interference; while single electrical parameter monitoring cannot identify optical characteristic faults such as brightness attenuation or color shift caused by aging of the encapsulation layer. Due to the lack of real-time correlation analysis and cross-verification of multi-source data (optical, electrical, and thermal), existing technologies struggle to accurately isolate environmental noise during dynamic playback and cannot achieve in-depth identification of pixel-level hidden faults and their failure mechanisms. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for identifying faults in LED displays, aiming to solve the problems mentioned in the background art.

[0005] This invention is implemented as follows: On one hand, a method for identifying faults in an LED display screen, the method comprising:

[0006] Synchronously acquire multimodal observation data of the LED display screen during the dynamic display process. The multimodal observation data includes optical images in the spatial domain, driving transient electrical signals in the time domain, and regional distribution heat maps.

[0007] The multimodal observation data is mapped to a unified physical pixel coordinate system, and cross-modal correlation alignment in the spatiotemporal dimension is performed.

[0008] Based on the optical image and the driving transient electrical signal, a photoelectric consistency verification model is constructed, and composite fault features of pixels are extracted.

[0009] The composite fault features are weighted and corrected based on the temperature rise trend of the heat map, and the final fault classification and diagnosis result is output.

[0010] As a further aspect of the present invention, the synchronous acquisition of multimodal observation data of the LED display screen during the dynamic display process specifically includes:

[0011] Use a high-speed vision sensor to acquire the integral brightness distribution data of the LED display screen during the current refresh cycle;

[0012] The rising edge voltage slope of the driver chip during the pulse width modulation (PWM) cycle is captured by the sampling circuit of the drive circuit.

[0013] Real-time temperature field data of the back panel of the screen is obtained using an infrared temperature sensing array;

[0014] The grayscale data of the control system output is extracted in real time and used as a logical reference benchmark for optical, electrical, and thermal feedback data.

[0015] As a further aspect of the present invention, mapping the multimodal observation data to a unified physical pixel coordinate system and performing cross-modal correlation alignment in the spatiotemporal dimension specifically includes:

[0016] Perform homography transformation on the optical image to achieve physical position mapping between display pixels and camera pixels;

[0017] The sampling points of the driving transient electrical signal are aligned with the exposure time window of the optical image by using the frame synchronization trigger signal;

[0018] The low-resolution temperature field data is fitted to a high-resolution pixel matrix using a spatial interpolation algorithm;

[0019] Construct a multimodal data fusion vector space indexed by pixel physical coordinates.

[0020] As a further aspect of the present invention, the step of constructing a photoelectric consistency verification model based on the optical image and the driving transient electrical signal, and extracting composite fault features of pixels, specifically includes:

[0021] The theoretical light intensity and theoretical impedance response of each pixel are calculated based on a logical reference.

[0022] Calculate the optical residual between the measured brightness and the theoretical light intensity, and the electrical residual between the measured voltage slope and the theoretical impedance;

[0023] Perform cross-correlation analysis of photoelectric residuals to identify photoelectric synchronization anomalies, single optical anomalies, and single electrical anomaly modes;

[0024] By fusing spatial contrast features and temporal waveform features, a composite fault feature characterizing the health status of a pixel is generated.

[0025] As a further aspect of the present invention, the step of combining the temperature rise trend of the heat map with the weight correction of the composite fault features to output the final fault classification and diagnosis result specifically includes:

[0026] Analyze the instantaneous temperature rise rate and local hot spot characteristics of the corresponding pixel regions in the distributed heat map;

[0027] If the temperature rise rate exceeds the preset threshold and is accompanied by an increase in electrical residual, the weight of the hardware short circuit fault determination will be increased.

[0028] By comparing composite fault characteristics with a multidimensional fault fingerprint database, display abnormalities caused by LED light decay, driver channel leakage, and external environmental interference can be distinguished.

[0029] The output includes a diagnostic report containing the physical coordinates of the fault, the failure type and mechanism, and the predicted remaining lifespan.

[0030] As a further aspect of the present invention, an LED display screen fault identification system is provided, the system comprising:

[0031] The multimodal information acquisition module is used to synchronously acquire raw feedback signals from different physical dimensions while the display screen is running.

[0032] A cross-modal space alignment module is used to achieve accurate matching of heterogeneous data in physical coordinates and time series.

[0033] The optoelectronic coupling analysis module is used to uncover deep differences in physical properties through virtual-real comparison and optoelectronic verification algorithms.

[0034] The integrated decision diagnosis module is used to introduce thermodynamics-assisted judgment to achieve accurate qualitative, quantitative and location-based fault identification.

[0035] As a further aspect of the present invention, the photoelectric coupling analysis module specifically includes:

[0036] The model prediction unit is used to generate the desired photoelectric physical feature vector based on the current displayed content;

[0037] The residual calculation unit is used to extract the multidimensional deviation between the measured physical quantity and the model prediction.

[0038] The feature fusion unit is used to identify correlational anomalies between photoelectric signals through a self-attention mechanism.

[0039] As a further aspect of the present invention, the comprehensive decision diagnosis module specifically includes:

[0040] A thermal compensation unit is used to perform thermodynamic reference correction on photoelectric deviation based on the regional temperature.

[0041] The fingerprint recognition unit is used to match optical, electrical, and thermal feature combinations under different failure mechanisms;

[0042] The diagnostic visualization unit is used to dynamically present fault attributes on the screen maintenance map.

[0043] This invention provides a fault identification method and system for LED displays. By using time-based hard synchronization and spatial affine alignment of heterogeneous data from multiple sources (optical, electrical, and thermal), a photoelectric consistency digital twin verification model unaffected by video content interference is constructed. Combined with a confidence correction mechanism based on thermodynamic temperature rise trends, this method overcomes the technical bottleneck of hidden faults being prone to false alarms and difficult to identify under high dynamic display conditions, achieving non-invasive, high-accuracy pixel-level online fault diagnosis. Attached Figure Description

[0044] Figure 1 This is the main flowchart of a fault identification method for LED displays.

[0045] Figure 2 This is a flowchart illustrating a method for identifying faults in LED displays, which involves constructing an optoelectronic consistency verification model based on the optical image and the driving transient electrical signal, and extracting composite fault features of pixels.

[0046] Figure 3 This is a flowchart illustrating how a fault identification method for LED displays combines the temperature rise trend of the distributed heat map to weight the composite fault features and outputs the final fault classification and diagnosis result.

[0047] Figure 4 This is a main structure diagram of an LED display screen fault identification system. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0050] The present invention provides a method and system for identifying faults in LED displays, which solves the technical problems in the background art.

[0051] like Figure 1 The diagram shown is a main flowchart of an LED display screen fault identification method according to an embodiment of the present invention. The LED display screen fault identification method includes:

[0052] Step S100: Synchronously acquire multimodal observation data of the LED display screen during the dynamic display process;

[0053] The multimodal observation data includes spatial domain optical images, temporal domain driving transient electrical signals, and regional distribution heatmaps.

[0054] Step S200: Map the multimodal observation data to a unified physical pixel coordinate system and perform cross-modal correlation alignment in the spatiotemporal dimension;

[0055] Step S300: Construct a photoelectric consistency verification model based on the optical image and the driving transient electrical signal, and extract composite fault features of pixels;

[0056] Step S400: Combine the temperature rise trend of the heat map to perform weight correction on the composite fault features, and output the final fault classification and diagnosis result;

[0057] In this embodiment, the system captures the operating data of the LED display screen in real time during normal video playback via a high-speed data bus. At this time, the FPGA main control unit, acting as a hardware coordination layer, synchronously initiates acquisition tasks in the three dimensions of light, electricity, and heat, ensuring the inherent temporal homogeneity of heterogeneous data. The CPU thread within the edge computing gateway is responsible for mapping data of different resolutions and field of view to a virtual two-dimensional coordinate system based on the LED physical pixel matrix, ensuring that each pixel possesses characteristic attributes in the three dimensions of light, electricity, and heat. Then, a photoelectric consistency verification model is constructed, incorporating theoretical voltage, theoretical current, and theoretical luminous flux. By comparing the residuals between the measured physical quantities and the theoretical predictions, false positive interference caused by changes in brightness within the video content itself is eliminated, accurately extracting the physical layer attenuation signal that reflects the true state of the hardware. Finally, thermodynamic temperature rise distribution is introduced as a key auxiliary criterion. The low-frequency hysteresis characteristics of temperature are used to weight the confidence of high-frequency photoelectric anomaly characteristics. For example, when a sudden change in electrical parameters occurs in a certain area but the temperature does not rise abnormally, the system will reduce the alarm weight of short circuit fault. Thus, in complex outdoor high-brightness and high-dynamic interference scenarios, the system outputs a highly robust and accurate final diagnostic result. The whole process does not require interruption of the business screen, realizing a truly seamless online health check.

[0058] In a preferred embodiment of the present invention, the synchronous acquisition of multimodal observation data of the LED display screen during the dynamic display process specifically includes:

[0059] Step S101: Use a high-speed vision sensor to acquire the integral brightness distribution data of the LED display screen during the current refresh cycle;

[0060] Step S102: Capture the rising edge voltage slope of the driver chip within the pulse width modulation (PWM) cycle through the sampling circuit of the drive circuit;

[0061] Step S103: Use an infrared temperature sensing array to acquire real-time temperature field data of the back panel of the screen;

[0062] Step S104: Extract the output grayscale data of the control system in real time as a logical reference benchmark for optical, electrical, and thermal feedback data.

[0063] In this embodiment, when acquiring visual image data, an industrial-grade high-frame-rate CMOS camera with a global shutter is deployed in front of the LED display screen. The camera's exposure time is strictly locked to an integer multiple of the display screen's PWM cycle, thereby completely eliminating moiré patterns and image tearing caused by high-frequency scanning and obtaining accurate pixel-level integrated brightness. For capturing the underlying electrical signals, considering that conventional ADCs cannot capture microsecond-level transient changes in drive signals, a high-speed analog-to-digital converter is externally connected to the constant current drive chip's readback pin of the receiving card. A trigger pulse strictly aligned with the rising edge of the PWM is generated by the phase-locked loop inside the FPGA, characterizing the voltage overshoot and slope features of the drive pin at the moment of conduction. For acquiring temperature field data, a wide-field-of-view infrared thermal imaging sensor array matrix is ​​distributed on the back panel of the enclosure, and a thermal distribution map including the environmental gradient background is reported in real time at a frequency of 4Hz via the I2C bus. To ensure time base consistency for all front-end data, a hardware data probe is embedded at the end of the video pipeline of the sending card to directly intercept the RGB source grayscale data stream currently being sent to each cabinet. The FPGA adds a unified 64-bit hardware timestamp to this batch of video streams, camera exposure completion signals, ADC sampling pulses, and thermal infrared data frames, which serves as the absolute logical reference benchmark for subsequent multimodal correlation analysis.

[0064] In a preferred embodiment of the present invention, mapping the multimodal observation data to a unified physical pixel coordinate system and performing cross-modal correlation alignment in the spatiotemporal dimension specifically includes:

[0065] Step S201: Perform homography transformation on the optical image to achieve physical position mapping between display pixels and camera pixels;

[0066] Step S202: Align the sampling points of the driving transient electrical signal with the exposure time window of the optical image using the frame synchronization trigger signal;

[0067] Step S203: Use a spatial interpolation algorithm to fit the low-resolution temperature field data to a high-resolution pixel matrix;

[0068] Step S204: Construct a multimodal data fusion vector space indexed by pixel physical coordinates.

[0069] In this embodiment, when addressing spatial misalignment issues, since industrial cameras typically have a certain installation tilt, a homography transformation algorithm based on feature point matching is employed. This algorithm extracts the positioning crosshairs at the four corners of the LED screen, calculates the perspective transformation matrix between the camera image plane and the physical plane of the LED display, and uses this matrix to perform inverse reprojection on the acquired distorted optical image. This allows the resampled image to be precisely bound to the physical LED beads at the pixel level. To address the temporal matching problem due to inconsistent sampling rates, a circular buffer is created in memory, using the rising edge of the frame synchronization trigger pulse as the alignment anchor point. The statistical mean and variance of the transient electrical signal sequence within the same exposure time window are extracted to complete the temporal downsampling alignment. Considering that the resolution of the infrared thermal imager is much lower than the LED bead density, a bicubic interpolation algorithm is used, combined with the partial differential equation of heat conduction from the aluminum heat sink on the back of the screen, to smoothly upsample and expand the dimension of the low-resolution original temperature field, generating a high-resolution temperature field of the same size as the LED bead array. After the above dimensionality reduction and alignment processes, a high-dimensional tensor space with physical coordinates as the key is instantiated in memory.

[0070] like Figure 2 As shown, in a preferred embodiment of the present invention, the step of constructing a photoelectric consistency verification model based on the optical image and the driving transient electrical signal, and extracting composite fault features of pixels specifically includes:

[0071] Step S301: Calculate the theoretical light intensity and theoretical impedance response of each pixel based on the logical reference.

[0072] Step S302: Calculate the optical residual between the measured brightness and the theoretical light intensity, and the electrical residual between the measured voltage slope and the theoretical impedance;

[0073] Step S303: Perform cross-correlation analysis of photoelectric residuals to identify photoelectric synchronization anomalies, single optical anomalies, and single electrical anomaly modes;

[0074] Step S304: Fuse spatial contrast features and temporal waveform features to generate composite fault features that characterize the health status of pixels.

[0075] It should be understood that during the model pre-setting stage, the Gamma correction curves of different batches of LED beads and the current-voltage characteristic curves of the driver chip are pre-fixed in non-volatile memory. When a grayscale instruction is received from the logic reference, the microprocessor quickly calculates the theoretical luminous intensity and theoretical on-resistance of the pixel in the current frame by looking up a table and using a linear interpolation algorithm. In the differential operation stage: on the one hand, the optical residual between the measured brightness of the camera and the theoretical light intensity is calculated; on the other hand, the electrical residual between the measured voltage rise slope of the ADC and the theoretical impedance is calculated, effectively filtering out global common-mode variables caused by video playback screen switching. When performing photoelectric cross-correlation analysis, the Pearson correlation coefficient is used to quantify the temporal linkage of the residuals. If a pixel shows a sudden drop in optical residual but the electrical residual approaches zero, it is determined that the screen surface is blocked by a foreign object; if the optical and electrical residuals undergo nonlinear distortion simultaneously, it is confirmed as physical damage at the die level. In the final feature encapsulation stage, the spatial contrast mutation features of surrounding pixels extracted based on the Laplacian operator are combined with the residual drift features in the time domain to reduce dimensions and output a series of composite fault feature vectors that are not affected by external lighting and display content.

[0076] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of weighting the composite fault features based on the temperature rise trend of the heat map and outputting the final fault classification and diagnosis result specifically includes:

[0077] Step S401: Analyze the instantaneous temperature rise rate and local hot spot characteristics of the corresponding pixel area in the distributed heat map;

[0078] Step S402: If the temperature rise rate exceeds the preset threshold and is accompanied by an increase in electrical residual, then increase the judgment weight of hardware short circuit fault.

[0079] Step S403: Compare the composite fault characteristics with the multidimensional fault fingerprint database to distinguish between display abnormalities caused by lamp bead light decay, drive channel leakage, and external environmental interference;

[0080] Step S404: Output a diagnostic report containing the physical coordinates of the fault, the failure type and mechanism, and the predicted remaining life.

[0081] In this embodiment, when thermodynamic auxiliary criteria are introduced, the diagnostic process continuously monitors the dynamic changes of the distributed heat map and uses a Kalman filter to estimate the instantaneous temperature rise rate of the corresponding pixel area over the past few frames to capture tiny hot spots caused by abnormal power consumption. In the weight correction stage, a set of adaptive fuzzy logic reasoning rules is built-in. For example, if the electrical residual indicates that the current of a certain driving channel is too large and the local temperature rise rate exceeds the preset safety threshold, the judgment weight for internal metallization layer short circuit or driver chip breakdown will be significantly increased; conversely, if the current is large but the temperature rise is slow, it is very likely due to normal impedance drift caused by excessively low ambient temperature, and the short circuit alarm will be automatically suppressed. In the fault type matching stage, relying on a pre-trained support vector machine classifier, the corrected composite fault features are sent to a multi-dimensional fault fingerprint database for hyperplane distance matching. This fingerprint database stores a massive number of aging attenuation and pin solder joint feature templates obtained through destructive experiments. At the output end of the diagnostic results, the successfully matched fault categories, degradation severity levels and physical coordinates are packaged into structured data packets and pushed to the UI interface of the monitoring center. The fault points are highlighted in the virtual maintenance topology map, and maintenance suggestions such as "abnormal overheating of module chips" are directly generated.

[0082] like Figure 4 As shown, in another preferred embodiment of the present invention, an LED display screen fault identification system is provided, the system comprising:

[0083] The multimodal information acquisition module 100 is used to synchronously acquire raw feedback signals from different physical dimensions while the display screen is running.

[0084] Cross-modal space alignment module 200 is used to achieve accurate matching of heterogeneous data in physical coordinates and time series.

[0085] The optoelectronic coupling analysis module 300 is used to uncover deep physical performance differences through virtual-real comparison and optoelectronic verification algorithms.

[0086] The integrated decision diagnosis module 400 is used to introduce thermodynamics-assisted judgment to achieve accurate qualitative, quantitative and location of faults.

[0087] In this embodiment, the multimodal information acquisition module 100, acting as a sensing probe, consists of various micro-sensor arrays embedded within the LED housing and a front-facing optical gimbal, responsible for capturing raw data at different timescales from nanoseconds to milliseconds. The cross-modal spatial alignment module 200, serving as the core hub of the data flow, is deployed in an FPGA accelerator card with large-capacity memory. Utilizing a pipelined parallel processing structure implemented with a hardware description language, it performs real-time spatiotemporal matrix alignment of heterogeneous data without consuming host CPU resources. The optoelectronic coupling analysis module 300, as the core algorithm process of the edge computing gateway, uses the floating-point computing power of the GPU to perform theoretical model calculations and cross-correlation calculations of massive residual tensors, extracting pure hardware performance anomaly signals. The comprehensive decision diagnosis module 400, deployed in the backend server, undertakes tasks such as thermal weight evaluation based on business logic and machine learning model inference.

[0088] In another preferred embodiment of the present invention, the photoelectric coupling analysis module 300 specifically includes:

[0089] The model prediction unit 301 is used to generate the desired photoelectric physical feature vector based on the current display content;

[0090] The residual calculation unit 302 is used to extract the multidimensional deviation between the measured physical quantity and the model prediction.

[0091] The feature fusion unit 303 is used to identify correlational anomalies between photoelectric signals through a self-attention mechanism.

[0092] In this embodiment, the model prediction unit 301 encapsulates a digital twin model. This model is input with prior parameters such as real-time grayscale playback content and factory batch aging rate, and outputs a virtual reference frame reflecting the theoretical state of the current screen frame through polynomial fitting. The residual calculation unit 302, as a high-performance matrix differencer, performs bitwise subtraction between the aligned measured photoelectric data and the virtual reference frame. Since the conventional high dynamic range image content exists synchronously in the theoretical model and measured data, the difference operation can perfectly cancel them out. The final residual data is entirely due to physical quantity deviations caused by hardware aging or electrical short circuits. The feature fusion unit 303 is responsible for deep mining of this batch of high-dimensional residual data. It adopts a self-attention mechanism based on the Transformer architecture, focusing on whether there is a cascade amplification effect in the photoelectric residuals of spatially adjacent pixels, and whether the impedance residual of a certain pixel in the time series exhibits periodic oscillations. Through this virtual-real comparison mechanism, weak early failure signals that were previously completely masked by complex video content are purified into significant statistical features.

[0093] In another preferred embodiment of the present invention, the comprehensive decision diagnosis module 400 specifically includes:

[0094] Thermal compensation unit 401 is used to perform thermodynamic reference correction on photoelectric deviation based on regional temperature.

[0095] The fingerprint recognition unit 402 is used to match optical, electrical, and thermal feature combination templates under different failure mechanisms;

[0096] The diagnostic visualization unit 403 is used to dynamically present fault attributes on the screen maintenance map.

[0097] In this embodiment, the thermal compensation unit 401 incorporates a semiconductor bandgap physics model. When a drastic fluctuation in substrate temperature is detected, this unit automatically performs thermodynamic reverse bias compensation on the previously extracted photoelectric residual baseline, significantly reducing the probability of misjudging physical breakdown due to increased semiconductor dark current caused by extreme high temperatures. The fingerprint recognition unit 402 employs a lightweight neural network deployed at the edge, projecting the thermally compensated feature vector into a high-dimensional feature space and measuring its distance from preset typical failure cluster centers. The cluster with the closest distance and a confidence level greater than a threshold is the finally identified fault mechanism. The diagnostic visualization unit 403 uses a modern front-end framework to develop a real-time inspection dashboard. After the backend confirms a fault, this unit pushes the diagnostic results to the browser via a long-connection protocol. On a vector map rendered according to the screen's physical wiring topology, the fault location is dynamically marked with thermal flashing points of different colors, and a digital work order containing inferences about the cause of failure is automatically generated.

[0098] The above embodiments of the present invention provide an LED display screen fault identification method and an LED display screen fault identification system. By using time-based hard synchronization and spatial affine alignment of heterogeneous data from multiple sources (optical, electrical, and thermal), a photoelectric consistency digital twin verification model unaffected by video content interference is constructed. Combined with a confidence correction mechanism based on thermodynamic temperature rise trends, the technical bottleneck of easily false alarms and difficulty in identifying hidden faults under high dynamic display conditions is solved, achieving non-invasive, high-accuracy pixel-level online fault diagnosis.

[0099] In order for the above methods and systems to operate smoothly, the system may include more or fewer components than those described above, or combine certain components, or different components, in addition to the various modules mentioned above. For example, it may include input / output devices, network access devices, buses, processors, and memory.

[0100] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (OPGs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the system, connecting various parts via various interfaces and lines.

[0101] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0102] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0103] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for fault identification of an LED display screen, characterized in that, The method includes: Synchronously acquire multimodal observation data of the LED display screen during the dynamic display process. The multimodal observation data includes optical images in the spatial domain, driving transient electrical signals in the time domain, and regional distribution heat maps. The multimodal observation data is mapped to a unified physical pixel coordinate system, and cross-modal correlation alignment in the spatiotemporal dimension is performed. Based on the optical image and the driving transient electrical signal, a photoelectric consistency verification model is constructed, and composite fault features of pixels are extracted. The composite fault features are weighted and corrected based on the temperature rise trend of the heat map, and the final fault classification and diagnosis result is output.

2. The LED display screen fault identification method according to claim 1, characterized in that, The synchronous acquisition of multimodal observation data of the LED display screen during the dynamic display process specifically includes: Use a high-speed vision sensor to acquire the integral brightness distribution data of the LED display screen during the current refresh cycle; The rising edge voltage slope of the driver chip during the pulse width modulation (PWM) cycle is captured by the sampling circuit of the drive circuit. Real-time temperature field data of the back panel of the screen is obtained using an infrared temperature sensing array; The grayscale data of the control system output is extracted in real time and used as a logical reference benchmark for optical, electrical, and thermal feedback data.

3. The LED display screen fault identification method according to claim 1, characterized in that, Mapping the multimodal observation data to a unified physical pixel coordinate system and performing cross-modal correlation alignment in the spatiotemporal dimension specifically includes: Perform homography transformation on the optical image to achieve physical position mapping between display pixels and camera pixels; The sampling points of the driving transient electrical signal are aligned with the exposure time window of the optical image by using the frame synchronization trigger signal; The low-resolution temperature field data is fitted to a high-resolution pixel matrix using a spatial interpolation algorithm; Construct a multimodal data fusion vector space indexed by pixel physical coordinates.

4. The LED display screen fault identification method according to claim 1, characterized in that, The step of constructing a photoelectric consistency verification model based on the optical image and the driving transient electrical signal, and extracting composite fault features of pixels, specifically includes: The theoretical light intensity and theoretical impedance response of each pixel are calculated based on a logical reference. Calculate the optical residual between the measured brightness and the theoretical light intensity, and the electrical residual between the measured voltage slope and the theoretical impedance; Perform cross-correlation analysis of photoelectric residuals to identify photoelectric synchronization anomalies, single optical anomalies, and single electrical anomaly modes; By fusing spatial contrast features and temporal waveform features, a composite fault feature characterizing the health status of a pixel is generated.

5. The LED display screen fault identification method according to claim 1, characterized in that, The step of combining the temperature rise trend of the heat map with the composite fault features for weight correction, and outputting the final fault classification and diagnosis result, specifically includes: Analyze the instantaneous temperature rise rate and local hot spot characteristics of the corresponding pixel regions in the distributed heat map; If the temperature rise rate exceeds the preset threshold and is accompanied by an increase in electrical residual, the weight of the hardware short circuit fault determination will be increased. By comparing composite fault characteristics with a multidimensional fault fingerprint database, display abnormalities caused by LED light decay, driver channel leakage, and external environmental interference can be distinguished. The output includes a diagnostic report containing the physical coordinates of the fault, the failure type and mechanism, and the predicted remaining lifespan.

6. A fault identification system for an LED display screen, characterized in that, The system, employing the LED display screen fault identification method as described in any one of claims 1-5, comprises: The multimodal information acquisition module is used to synchronously acquire raw feedback signals from different physical dimensions while the display screen is running. A cross-modal space alignment module is used to achieve accurate matching of heterogeneous data in physical coordinates and time series. The optoelectronic coupling analysis module is used to uncover deep differences in physical properties through virtual-real comparison and optoelectronic verification algorithms. The integrated decision diagnosis module is used to introduce thermodynamics-assisted judgment to achieve accurate qualitative, quantitative and location-based fault identification.

7. The LED display screen fault identification system according to claim 6, characterized in that, The photoelectric coupling analysis module specifically includes: The model prediction unit is used to generate the desired photoelectric physical feature vector based on the current displayed content; The residual calculation unit is used to extract the multidimensional deviation between the measured physical quantity and the model prediction. The feature fusion unit is used to identify correlational anomalies between photoelectric signals through a self-attention mechanism.

8. The LED display screen fault identification system according to claim 6, characterized in that, The comprehensive decision diagnosis module specifically includes: A thermal compensation unit is used to perform thermodynamic reference correction on photoelectric deviation based on the regional temperature. The fingerprint recognition unit is used to match optical, electrical, and thermal feature combinations under different failure mechanisms; The diagnostic visualization unit is used to dynamically present fault attributes on the screen maintenance map.

Images