Health state prediction method for 8K display device

Abstract

The invention provides a health state prediction method for an 8K display device, and relates to the technical field of display health prediction.The health state prediction method comprises the steps that a multi-modal feature vector is extracted from each pixel node of a multi-modal frequency spectrum image, a mutual information feature vector of the multi-modal feature vector is calculated, and subtle changes of cooperative work among different physical attributes in the nodes are captured; on the basis, each pixel node is further regarded as a node in a pixel node grid, the connection weight is quantized by calculating the norm of a mutual information feature vector of the pixel node and obtaining the similarity between adjacent nodes, then the attenuation rate of the connection weight of a continuous time window is calculated, the distribution entropy value of the attenuation rate is analyzed, and a stability degradation entropy index is obtained; trivial connection change information is condensed into a quantitative index representing the overall orderliness decline degree of the system, then time sequence prediction is carried out on an entropy index historical sequence, the expected time of the entropy index historical sequence reaching the dynamic control upper limit is calculated, and the crossing from current state monitoring to future trend prediction is achieved.

Description

Technical Field

[0001] This invention relates to the field of display health prediction technology, and in particular to a method for predicting the health status of an 8K display device. Background Technology

[0002] With the rapid development of ultra-high resolution display technology, 8K display devices have been widely used in various fields. Their long-term reliability and stability are crucial. However, 8K displays contain tens of millions or even hundreds of millions of pixels. Their structural complexity and huge pixel scale make traditional health monitoring methods based on single performance parameter threshold alarms or simple image processing difficult to cope with. Existing technologies usually assess the health status of the display by monitoring overall brightness uniformity, chromaticity coordinates, or detecting isolated dead pixels. These methods have significant limitations. They can only issue alarms when performance degradation has reached a certain level and faults have already appeared. They are reactive and cannot achieve early prediction and warning of faults. More importantly, the failure of display devices often begins with deep physical and chemical changes such as internal material aging, performance degradation of driving circuits, or failure of micro-connections. These changes initially manifest as instability in the synergistic operation between different physical properties, and eventually gradually manifest as visible display defects. These macroscopic performance indicators cannot reveal early, localized performance degradation inside the display, let alone quantify the decline trend of health status.

[0003] Therefore, it is necessary to provide a health status prediction method for 8K display devices to solve the above-mentioned technical problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for predicting the health status of 8K display devices, achieving the beneficial effect of efficiently and accurately predicting the health status of 8K displays.

[0005] This invention provides a method for predicting the health status of an 8K display device, comprising: S1: Based on a preset detection cycle, periodically and synchronously acquire multimodal spectrum images of the display device under a uniform test screen, and register and align the multimodal spectrum images to a unified pixel node grid; S2: Extract the multimodal feature vector of the multimodal spectrum image corresponding to each pixel node in the pixel node grid, and calculate the mutual information feature vector of each pixel node based on multiple consecutive preset detection cycles within a preset sliding time window. S3: Based on the mutual information feature vector of each pixel node, calculate the connection weight between any two adjacent pixel nodes in the pixel node grid to obtain the connection weight matrix; S4: Based on the connection weight matrix obtained within two adjacent preset sliding time windows, calculate the connection weight decay rate and obtain the weight decay rate matrix. S5: Based on the weight decay rate matrix, calculate the stability degradation entropy index, combine multiple consecutive stability degradation entropy indices into a degradation entropy index sequence, and calculate the current dynamic control upper limit, current degradation rate and current degradation acceleration based on the degradation entropy index sequence, and generate a sequence of entropy index prediction values. S6: Calculate the expected time window for the entropy index prediction value sequence to reach the current dynamic control upper limit, calculate the remaining stable time based on the expected time window, and generate a health prediction report by combining the current degradation rate and the current degradation acceleration.

[0006] Preferably, in step S2, the calculation of the mutual information feature vector includes: For each pixel node, calculate the instantaneous mutual information between any two different modal features in its multimodal feature vector within a preset sliding time window of time-series data, and combine all the instantaneous mutual information values ​​into the mutual information feature vector of that pixel node.

[0007] Preferably, in step S3, the calculation of the connection weights includes: For each pixel node, calculate the L2 norm value of its mutual information feature vector, and use the L2 norm value as the comprehensive cooperative strength of the pixel node. Calculate the Gaussian similarity between the combined collaborative strength values ​​of each pair of adjacent pixel nodes, and use the Gaussian similarity value as the connection weight between the pair of adjacent pixel nodes.

[0008] Preferably, the bandwidth parameter used in calculating the Gaussian similarity is determined based on the standard deviation of the comprehensive cooperative strength values ​​of all pixel nodes within multiple consecutive preset detection periods in a preset sliding time window.

[0009] Preferably, in step S5, the probability distribution of the values ​​of all elements in the weight decay rate matrix is ​​statistically analyzed, and the information entropy of the probability distribution is calculated as the stability degradation entropy index.

[0010] Preferably, in step S5, the calculation steps for the current dynamic control upper limit include: Calculate the mean and standard deviation of all stability degradation entropy index values ​​in the degradation entropy index sequence; The current dynamic control upper limit is obtained by multiplying the average value plus the standard deviation by a preset proportional coefficient.

[0011] Preferably, in step S5, the calculation steps for the current degradation rate and the current degradation acceleration include: By applying the multinomial regression algorithm, the degradation entropy index sequence is fitted with an independent function of at least second order to obtain the entropy index trend fitting function of the degradation entropy index sequence. Calculate the first and second derivatives of the entropy exponential trend fitting function, and use them as the current degradation rate and current degradation acceleration, respectively.

[0012] Preferably, step S6 further includes triggering a trend deterioration warning signal when the average of multiple consecutive entropy index prediction values ​​in the entropy index prediction value sequence exceeds the current dynamic control upper limit.

[0013] Preferably, step S6 further includes the following steps: From the weight decay rate matrix corresponding to the triggering trend deterioration warning signal, the pixel nodes corresponding to the connection weights whose weight decay rate values ​​are higher than the preset high decay threshold are selected to form the initial abnormal node set. A spatial density-based clustering algorithm is applied to the initial set of anomalous nodes to identify subsets of anomalous nodes that are closely adjacent in the two-dimensional spatial coordinates of the pixel node grid, and each subset of anomalous nodes is taken as an anomalous node cluster. The abnormal nodes contained in the abnormal node cluster are mapped to the physical display area of ​​the display device according to their two-dimensional spatial coordinates in the pixel node grid, thus forming a potential fault area.

[0014] Preferably, the spatial density-based clustering algorithm is the DBSCAN algorithm, in which the neighborhood radius parameter is pre-set based on the physical resolution of the display corresponding to the pixel node grid and the typical size of the historical potential fault area.

[0015] Compared with related technologies, the health status prediction method for 8K display devices provided by this invention has the following beneficial effects: This invention first establishes a solid foundation for multi-dimensional physical information fusion analysis by simultaneously acquiring multi-modal spectral images of visible light, short-wave infrared, and ultraviolet fluorescence, and performing high-precision registration. Then, it extracts multi-modal feature vectors from each pixel node and calculates their mutual information feature vectors. This step can keenly capture subtle changes in the collaborative work between different physical properties within a node, thereby revealing potential early signs of performance degradation at the microscopic level. Furthermore, each pixel node is treated as a node in a pixel node grid. By calculating the norm of its mutual information feature vector and obtaining the Gaussian similarity between adjacent nodes, the connection weights are quantified, enabling… The health assessment object has been elevated from isolated pixels to an interconnected functional system. Then, the decay rate of the connection weights in a continuous time window is calculated and its distribution entropy value is analyzed to obtain the stability degradation entropy index. This transforms fragmented connection change information into a single quantitative indicator that characterizes the degree of overall system order degradation, improving the efficiency and intuitiveness of the state assessment. Subsequently, by performing time series prediction on the historical entropy index and calculating the expected time to reach the dynamic control upper limit, a leap from current state monitoring to future trend prediction is achieved. The provided predicted remaining stable time provides a key time basis for formulating accurate predictive maintenance plans. Attached Figure Description

[0016] Figure 1 This is a flowchart of a health status prediction method for an 8K display device according to the present invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the drawings, not all structures. Moreover, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0018] It should also be noted that, for ease of description, the accompanying drawings show only the parts relevant to the invention and not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as being processed sequentially, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but it may also have additional steps not included in the drawings. The process may correspond to a method, function, procedure, subroutine, subprogram, etc.

[0019] Example 1 A method for predicting the health status of an 8K display device, in its specific implementation, such as... Figure 1 As shown, a flowchart of a health status prediction method for an 8K display device according to the present invention is illustrated, including: Step S1: Based on the preset detection cycle, periodically and synchronously acquire multimodal spectrum images of the display device under a uniform test screen, and register and align the multimodal spectrum images to a unified pixel node grid.

[0020] In the specific implementation process, multimodal spectrum images presented by the display under a uniform test screen are periodically and synchronously acquired through a preset detection cycle. These multimodal spectrum images are then precisely registered and aligned onto a unified pixel node grid. The preset detection cycle is set based on the actual usage frequency and expected lifespan of the display device; for example, it is set to trigger acquisition every 1 hour to ensure the continuity and representativeness of data acquisition. The uniform test screen is selected as a standard test pattern, for example, including but not limited to the grayscale card recommended by the International Commission on Illumination, to ensure that the display device is in a stable and reproducible working state, thereby eliminating the interference of content changes on the acquired data. The multimodal spectrum images include, but are not limited to, visible light spectrum images, short-wave infrared images, and ultraviolet fluorescence images. Visible light spectrum images are acquired by a high-resolution visible light camera to capture the brightness distribution and color uniformity of the display device. Short-wave infrared images are acquired by a short-wave infrared camera to detect the temperature distribution and thermal radiation characteristics of the display device surface, thereby indirectly reflecting the working status and heat dissipation efficiency of the driving circuit. Ultraviolet fluorescence images... Fluorescence images are acquired using an ultraviolet-excited camera to detect fluorescence effects caused by aging or potential defects in the optical materials of display devices. The synchronous acquisition of these images is achieved by synchronizing all cameras via a hardware trigger signal, ensuring that the multimodal images acquired at the same time are perfectly aligned in time, avoiding registration errors caused by acquisition time differences. After acquisition, the multimodal images need to be registered and aligned to a unified pixel node grid. This pixel node grid is a digital coordinate system that corresponds one-to-one with the physical pixel array of the 8K display device, with each pixel node representing the center position of a physical pixel. The registration process first uses image processing algorithms, including but not limited to scale-invariant feature transformation and accelerated robust feature algorithms, to extract key feature points from each modal image. Then, these feature points are used to calculate spatial transformation parameters between images, thereby correcting the distortion of images from different modalities to the same coordinate system, achieving pixel-level precise alignment. The registered image data is mapped to the pixel node grid, so that each grid node contains spectral information from the multimodal images, laying the data foundation for subsequent feature extraction and health analysis.

[0021] Step S2: Extract the multimodal feature vector of the multimodal spectrum image corresponding to each pixel node in the pixel node grid, and calculate the mutual information feature vector of each pixel node based on multiple consecutive preset detection periods within a preset sliding time window.

[0022] Specifically, in step S2, the calculation steps for the mutual information feature vector include: For each pixel node, calculate the instantaneous mutual information between any two different modal features in its multimodal feature vector within a preset sliding time window of time-series data, and combine all the instantaneous mutual information values ​​into the mutual information feature vector of that pixel node.

[0023] In the specific implementation process, the multimodal feature vector corresponding to each pixel node is extracted from the multimodal spectrum image that has been registered and aligned to a unified pixel node grid. The multimodal feature vector includes, but is not limited to, the brightness value and chromaticity coordinates extracted from the visible light spectrum image, the temperature value obtained from the short-wave infrared image, and the fluorescence intensity value extracted from the ultraviolet fluorescence image. These specific physical quantities constitute the original feature representation of each node in the multimodal spectrum. Then, the mutual information calculation stage is entered. For each specific pixel node, the time series data of its multimodal feature vector over multiple consecutive detection cycles within a preset sliding time window is taken. For example, thirty consecutive cycles are taken as the length of a preset sliding time window, and the sliding step size is the cycle interval of the preset detection cycle. Data is collected once per cycle, and any value in the multimodal feature vector of the pixel node is calculated. The instantaneous mutual information between time-series data sequences corresponding to two different modal features is exemplified by calculating the mutual information between pairs of modal features of each pixel node within thirty periods. This includes, but is not limited to, calculating the instantaneous mutual information between the brightness sequence and the temperature sequence, the brightness sequence and the fluorescence intensity sequence, and the temperature sequence and the fluorescence intensity sequence. For each pixel node, the instantaneous mutual information values ​​calculated from all pairs of modal feature pairs are combined into a new vector in a preset order, which is the mutual information feature vector of that pixel node. This mutual information feature vector comprehensively characterizes the dynamic correlation strength of various physical processes such as photothermal and electrical material properties within the pixel node in the time dimension. Finally, a corresponding mutual information feature vector is generated for each node in the entire pixel node grid, providing a basis for subsequent analysis of the functional connections between nodes.

[0024] Step S3: Based on the mutual information feature vector of each pixel node, calculate the connection weight between any two adjacent pixel nodes in the pixel node grid to obtain the connection weight matrix.

[0025] Specifically, in step S3, the calculation of the connection weights includes: For each pixel node, calculate the L2 norm value of its mutual information feature vector, and use the L2 norm value as the comprehensive cooperative strength of the pixel node. Calculate the Gaussian similarity between the combined collaborative strength values ​​of each pair of adjacent pixel nodes, and use the Gaussian similarity value as the connection weight between the pair of adjacent pixel nodes.

[0026] Specifically, the bandwidth parameter used to calculate Gaussian similarity is determined based on the standard deviation of the comprehensive cooperative strength values ​​of all pixel nodes within multiple consecutive preset detection periods in a preset sliding time window.

[0027] In the specific implementation process, based on the mutual information feature vector of each pixel node, the L2 norm of the mutual information feature vector of each pixel node in the pixel node grid is calculated. This is calculated as the square root of the sum of squares of all elements in the mutual information feature vector. This L2 norm value is a scalar and is defined as the comprehensive cooperative strength of that pixel node. Next, the connection weight between any pair of spatially adjacent pixel nodes in the pixel node grid is calculated using a Gaussian similarity function. Specifically, for a pair of adjacent pixel nodes, their respective comprehensive cooperative strength values ​​are substituted into the Gaussian similarity function. The formula for the Gaussian similarity function is: ,in, Represents pixel nodes and pixel nodes Connection weights between them and Representing pixel nodes and pixel nodes The overall synergy strength value, The bandwidth parameter of the Gaussian function affects the scale sensitivity of similarity calculation. This bandwidth parameter is not a fixed value but is dynamically determined based on the overall standard deviation of the comprehensive cooperative strength values ​​of all pixel nodes in multiple consecutive detection cycles within the current preset sliding time window. This allows the bandwidth parameter to adapt to the overall dispersion of the cooperative strength of the current network nodes, thus making the quantization of connection weights more robust. The above Gaussian similarity calculation is repeated for each pair of adjacent nodes, and the result is used as the connection weight of that node pair. Finally, the connection weights of all node pairs are systematically organized into a connection weight matrix. This connection weight matrix fully characterizes the strength of the functional connections between spatially adjacent nodes in the entire display pixel network under a specific time window, thus laying the foundation for subsequent analysis of the dynamic decay of network connection strength.

[0028] Step S4: Based on the connection weight matrix obtained within two adjacent preset sliding time windows, calculate the connection weight decay rate to obtain the weight decay rate matrix.

[0029] In the specific implementation process, firstly, two connection weight matrices calculated within two adjacent preset sliding time windows are obtained. These two connection weight matrices represent the functional connection strength between all adjacent node pairs in the entire pixel node grid under the previous preset time window and the current preset sliding time window, respectively. The current preset sliding time window refers to the preset sliding time window that includes the latest acquired multimodal spectral image data. Subsequently, for the connection weight value of each element at the same position in the connection weight matrix under the two consecutive preset sliding time windows, its normalized attenuation rate is calculated. Specifically, the calculation method is to subtract the connection weight value of the current preset sliding time window from the connection weight value of the previous preset sliding time window. In addition to the connection weight values ​​within a preset sliding time window, the weight decay rate of the connection weight matrix at that element position is obtained. A positive weight decay rate indicates that the connection weight has decayed, and the larger the value, the more severe the decay. If the weight decay rate is negative, it indicates that the connection strength has actually increased. The reasons include system fluctuations and measurement noise. The weight decay rate of each element in all connection weight matrices is calculated and arranged according to their corresponding positions in the connection weight matrices to form a new matrix, namely the weight decay rate matrix. Each element of the weight decay rate matrix accurately records the performance change rate of the connection weight over time, providing a direct data basis for the next step of statistically analyzing the overall degradation of network connection stability from a global perspective.

[0030] Step S5: Based on the weight decay rate matrix, calculate the stability degradation entropy index, combine multiple consecutive stability degradation entropy indices into a degradation entropy index sequence, and calculate the current dynamic control upper limit, current degradation rate and current degradation acceleration based on the degradation entropy index sequence, and generate an entropy index prediction value sequence.

[0031] Specifically, in step S5, the probability distribution of the values ​​of all elements in the weight decay rate matrix is ​​statistically analyzed, and the information entropy of the probability distribution is calculated as the stability degradation entropy index.

[0032] Specifically, in step S5, the calculation steps for the current dynamic control upper limit include: Calculate the mean and standard deviation of all stability degradation entropy index values ​​in the degradation entropy index sequence; The current dynamic control upper limit is obtained by multiplying the average value plus the standard deviation by the preset proportional coefficient.

[0033] Specifically, in step S5, the calculation steps for the current degradation rate and the current degradation acceleration include: By applying the multinomial regression algorithm, the degradation entropy index sequence is fitted with an independent function of at least second order to obtain the entropy index trend fitting function of the degradation entropy index sequence. Calculate the first and second derivatives of the entropy exponential trend fitting function, and use them as the current degradation rate and current degradation acceleration, respectively.

[0034] In the specific implementation process, firstly, the probability distribution of the attenuation rate values ​​of all connection weights in the weight attenuation rate matrix is ​​statistically analyzed, and the stability degradation entropy index is obtained by calculating the information entropy of the probability distribution. The stability degradation entropy index reflects the overall disorder of connection attenuation between pixel nodes. Subsequently, multiple consecutively calculated stability degradation entropy indices are combined into a degradation entropy index sequence in chronological order. Based on the degradation entropy index sequence, the current dynamic control upper limit is calculated. Specifically, the average and standard deviation of all entropy index values ​​in the degradation entropy index sequence are calculated. The average plus the standard deviation is then multiplied by a preset proportional coefficient to obtain the dynamic control upper limit. The preset proportional coefficient is an empirical parameter used to adjust the warning sensitivity. For example, the number is set within the range of two to three based on historical data. Its function is to define a statistical tolerance boundary relative to the historical normal fluctuation range. At the same time, time series prediction algorithms, including but not limited to autoregressive integral moving average models, are applied to fit the degradation entropy index sequence to generate an entropy index prediction value sequence to predict future trends. In addition, in order to quantify the degradation dynamics, a multinomial regression algorithm is applied to fit the degradation entropy index sequence with a function of at least second order to obtain the entropy index trend fitting function. The first and second derivative values ​​of the entropy index trend fitting function at the current time point are calculated as the current degradation rate and the current degradation acceleration, respectively, to characterize the speed and acceleration of change.

[0035] Step S6: Calculate the expected time window for the entropy index prediction value sequence to reach the current dynamic control upper limit, calculate the predicted remaining stabilization time based on the expected time window, and generate a health prediction report by combining the current degradation rate and the current degradation acceleration.

[0036] Specifically, step S6 further includes triggering a trend deterioration warning signal when the average of multiple consecutive entropy index prediction values ​​in the entropy index prediction value sequence exceeds the current dynamic control upper limit.

[0037] Specifically, step S6 also includes the following steps: From the weight decay rate matrix corresponding to the triggering trend deterioration warning signal, the pixel nodes corresponding to the connection weights whose weight decay rate values ​​are higher than the preset high decay threshold are selected to form the initial abnormal node set. A spatial density-based clustering algorithm is applied to the initial set of anomalous nodes to identify subsets of anomalous nodes that are closely adjacent in the two-dimensional spatial coordinates of the pixel node grid, and each subset of anomalous nodes is taken as an anomalous node cluster. The abnormal nodes contained in the abnormal node cluster are mapped to the physical display area of ​​the display device according to their two-dimensional spatial coordinates in the pixel node grid, thus forming a potential fault area.

[0038] Specifically, the clustering algorithm based on spatial density is the DBSCAN algorithm. The neighborhood radius parameter in the DBSCAN algorithm is pre-set based on the physical resolution of the display corresponding to the pixel node grid and the typical size of the historical potential fault area.

[0039] In the specific implementation process, the entropy index prediction value sequence is sequentially traversed. During the traversal, the first data point in the entropy index prediction value sequence whose single entropy index prediction value is greater than or equal to the current dynamic control upper limit is identified and recorded as the initial data point. The traversal continues to multiple consecutive data points after the initial data point. For example, the traversal continues to the two data points after the initial data point, and the entropy index prediction values ​​of these two data points are summed with the entropy index prediction value of the initial data point to calculate the arithmetic mean. If the arithmetic mean exceeds the current dynamic control upper limit, a trend deterioration warning signal is immediately triggered, indicating that the system health status shows a statistical trend of accelerated deterioration. At the same time, the future time corresponding to the initial data point is determined as the expected time window. If no initial data point is found in the entropy index prediction value sequence, the expected time window is extrapolated using linear interpolation based on the trend at the end of the sequence. Subsequently, the remaining stable time is predicted based on this expected time window. Furthermore, from the weight decay rate matrix corresponding to the triggering of the trend deterioration warning, connection weights with a weight decay rate higher than the preset high decay threshold are selected. The pixel nodes corresponding to these abnormal connections are identified to form an initial abnormal node set. Then, a spatial density-based clustering algorithm, DBSCAN algorithm, is applied to this node set. The steps for setting the key parameter neighborhood radius are as follows: First, the physical size of the display device and the resolution of the pixel node grid are obtained to calculate the physical size corresponding to a single pixel node. Then, the typical physical size of the potential fault area is obtained from the historical fault database. The typical physical size is divided by the physical size of a single pixel node to convert it into a typical size in units of pixel nodes. The typical size is then multiplied by a preset scaling factor to obtain the final neighborhood radius parameter. This parameter is used to perform DBSCAN clustering on the initial abnormal node set to identify a subset of abnormal nodes that are closely adjacent in the two-dimensional spatial coordinates of the pixel node grid. Each subset is defined as an abnormal node cluster. Finally, these clusters are mapped to the physical display area of ​​the display to form a potential fault area. Finally, the system integrates all information to generate a health prediction report. The report includes warning status indicators, predicted remaining stabilization time, current degradation rate and degradation acceleration values ​​and directions, detailed location and range of potential fault areas, risk level classification based on remaining time and degradation acceleration, and corresponding maintenance decision recommendations, thus comprehensively reflecting the health status and future risks of the display device.

[0040] The working principle of the health status prediction method for 8K display devices provided by this invention is as follows: This invention, based on a preset detection cycle, synchronously acquires multimodal spectral images of a display device under a uniform test screen, including visible light spectrum, short-wave infrared, and ultraviolet fluorescence images. These images are precisely registered onto a unified pixel node grid, establishing a feature vector of multidimensional physical quantities for each physical pixel. Subsequently, based on data from multiple consecutive detection cycles within a preset sliding time window, the instantaneous mutual information between different modal features within each pixel node is calculated, thereby generating a mutual information feature vector reflecting the collaborative working state of photothermal and electrophysiological processes within the pixel. Next, the comprehensive collaborative strength of the mutual information feature vectors of each pixel node is calculated, and the Gaussian similarity of the comprehensive collaborative strength between adjacent pixel nodes is further calculated to quantify their functional connection weights, thus forming a connection weight matrix that characterizes the functional association network between pixels of the display device. The system then compares the connection weight matrices of two adjacent time windows to calculate... The normalized decay rate of each connection weight is calculated to obtain a weight decay rate matrix, which captures the dynamic degradation behavior of the functional network. Based on the weight decay rate matrix, the information entropy of the numerical distribution of all its elements is calculated to obtain the stability degradation entropy index, which quantifies the overall disorder of network connection decay. After constructing a sequence of multiple continuous entropy indices, on the one hand, the mean and standard deviation of the sequence are calculated and combined with a preset proportional coefficient to determine the dynamic control upper limit. On the other hand, a time series prediction algorithm is applied to generate an entropy index prediction value sequence to predict future trends. At the same time, the current degradation rate and acceleration are calculated by performing polynomial regression fitting on the historical entropy index sequence to characterize the deterioration trend and urgency of the health status. Finally, the expected time for the entropy index prediction value sequence to reach the dynamic control upper limit is calculated to obtain the predicted remaining stable time. The degradation rate and acceleration are then integrated to generate a comprehensive health prediction report, realizing a complete closed loop from microscopic physical signal perception to macroscopic system health trend prediction.

[0041] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0042] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0043] It should also be noted that 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 process, method, article, or apparatus. Unless otherwise specified, 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 that element.

Claims

1. A method for predicting the health status of an 8K display device, characterized in that, The health status prediction method includes: S1: Based on a preset detection cycle, periodically and synchronously acquire multimodal spectrum images of the display device under a uniform test screen, and register and align the multimodal spectrum images to a unified pixel node grid; S2: Extract the multimodal feature vector of the multimodal spectrum image corresponding to each pixel node in the pixel node grid, and calculate the mutual information feature vector of each pixel node based on multiple consecutive preset detection cycles within a preset sliding time window. S3: Based on the mutual information feature vector of each pixel node, calculate the connection weight between any two adjacent pixel nodes in the pixel node grid to obtain the connection weight matrix; S4: Based on the connection weight matrix obtained within two adjacent preset sliding time windows, calculate the connection weight decay rate and obtain the weight decay rate matrix. S5: Based on the weight decay rate matrix, calculate the stability degradation entropy index, combine multiple consecutive stability degradation entropy indices into a degradation entropy index sequence, and calculate the current dynamic control upper limit, current degradation rate and current degradation acceleration based on the degradation entropy index sequence, and generate a sequence of entropy index prediction values. S6: Calculate the expected time window for the entropy index prediction value sequence to reach the current dynamic control upper limit, calculate the remaining stable time based on the expected time window, and generate a health prediction report by combining the current degradation rate and the current degradation acceleration.

2. The method for predicting the health status of an 8K display device according to claim 1, characterized in that, In step S2, the calculation steps for the mutual information feature vector include: For each pixel node, calculate the instantaneous mutual information between any two different modal features in its multimodal feature vector within a preset sliding time window of time-series data, and combine all the instantaneous mutual information values ​​into the mutual information feature vector of that pixel node.

3. The method for predicting the health status of an 8K display device according to claim 2, characterized in that, In step S3, the calculation of connection weights includes: For each pixel node, calculate the L2 norm value of its mutual information feature vector, and use the L2 norm value as the comprehensive cooperative strength of the pixel node. Calculate the Gaussian similarity between the combined collaborative strength values ​​of each pair of adjacent pixel nodes, and use the Gaussian similarity value as the connection weight between the pair of adjacent pixel nodes.

4. The health status prediction method for an 8K display device according to claim 3, characterized in that, The bandwidth parameter used to calculate the Gaussian similarity is determined based on the standard deviation of the comprehensive cooperative strength values ​​of all pixel nodes within multiple consecutive preset detection periods in a preset sliding time window.

5. The health status prediction method for an 8K display device according to claim 4, characterized in that, In step S5, the probability distribution of the values ​​of all elements in the weight decay rate matrix is ​​statistically analyzed, and the information entropy of the probability distribution is calculated as the stability degradation entropy index.

6. The health status prediction method for an 8K display device according to claim 5, characterized in that, In step S5, the calculation steps for the current dynamic control upper limit include: Calculate the mean and standard deviation of all stability degradation entropy index values ​​in the degradation entropy index sequence; The current dynamic control upper limit is obtained by multiplying the average value plus the standard deviation by a preset proportional coefficient.

7. The health status prediction method for an 8K display device according to claim 6, characterized in that, In step S5, the calculation steps for the current degradation rate and the current degradation acceleration include: By applying the multinomial regression algorithm, the degradation entropy index sequence is fitted with an independent function of at least second order to obtain the entropy index trend fitting function of the degradation entropy index sequence. Calculate the first and second derivatives of the entropy exponential trend fitting function, and use them as the current degradation rate and current degradation acceleration, respectively.

8. The method for predicting the health status of an 8K display device according to claim 7, characterized in that, Step S6 also includes triggering a trend deterioration warning signal when the mean of multiple consecutive entropy index prediction values ​​in the entropy index prediction value sequence exceeds the current dynamic control upper limit.

9. The health status prediction method for an 8K display device according to claim 8, characterized in that, Step S6 also includes the following steps: From the weight decay rate matrix corresponding to the triggering trend deterioration warning signal, the pixel nodes corresponding to the connection weights whose weight decay rate values ​​are higher than the preset high decay threshold are selected to form the initial abnormal node set. A spatial density-based clustering algorithm is applied to the initial set of anomalous nodes to identify subsets of anomalous nodes that are closely adjacent in the two-dimensional spatial coordinates of the pixel node grid, and each subset of anomalous nodes is taken as an anomalous node cluster. The abnormal nodes contained in the abnormal node cluster are mapped to the physical display area of ​​the display device according to their two-dimensional spatial coordinates in the pixel node grid, thus forming a potential fault area.

10. The health status prediction method for an 8K display device according to claim 8, characterized in that, The spatial density-based clustering algorithm is the DBSCAN algorithm. In the DBSCAN algorithm, the neighborhood radius parameter is pre-set based on the physical resolution of the display corresponding to the pixel node grid and the typical size of the historical potential fault area.

Images