Curved screen display method and device

Abstract

The invention relates to the technical field of curved screen fault diagnosis and compensation, and discloses a curved screen display method and device. The method comprises the following steps: capturing an original display signal flow containing a pixel signal, a driving time sequence and an embedded sensor signal, and generating a three-dimensional spatio-temporal data volume after synchronous alignment and standardized packaging; a dynamic anomaly perception map is constructed based on the data body, nodes of the map are composed of visual attributes and logic state units, and edges of the map are defined by physical quantity change relations. Next, nodes in the map are associated and mapped to a curved screen physical model, abnormal association strength is calculated, and a screen physical area and a signal driving channel corresponding to an abnormal source node are backtracked and positioned according to the abnormal association strength; and generating a targeted pixel signal compensation sequence and a driving parameter compensation sequence according to a positioning result. According to the method, accurate diagnosis and directional compensation of display abnormity from a phenomenon to a physical source and a circuit channel are realized.

Description

Technical Field

[0001] This invention relates to the field of curved screen fault diagnosis and compensation technology, specifically to a curved screen display method and device. Background Technology

[0002] In the field of curved screen displays, ensuring image uniformity and stability hinges on real-time monitoring and compensation for display anomalies. Existing solutions typically employ a discrete processing approach: an image processing engine analyzes the input video signal and detects image defects by identifying pixel block features; a timing controller generates scanning and data signals based on preset driving parameters; and sensors integrated within the screen independently report environmental data such as curvature, temperature, or pressure. Each system operates within its own logical closed loop, with limited data exchange or simple threshold linkage only at the final stage.

[0003] The drawback of this processing model is that it leads to a deep disconnect between the "sensory performance," "driving logic," and "physical causes" of display anomalies. When complex anomalies such as flickering, color shift, or ghosting occur in localized areas of the screen, the system struggles to establish an effective causal chain. Image processing can locate abnormal pixel areas, but it cannot determine whether the anomaly originates from an output misalignment of a specific channel in the source driver chip, a drift in localized TFT characteristics caused by dynamic bending of the screen, or a combination of both. Due to the lack of quantitative analysis of cross-domain correlations, subsequent compensation often becomes a superficial, general adjustment, such as patching the overall brightness or gamma value of the abnormal area. This is not only inefficient but may also cause new display problems by failing to correct underlying driving errors.

[0004] There is a need for a technology that can fundamentally break down the analytical barriers between visual signals, driving logic, and physical states. The goal of this invention is to solve the fundamental problem that existing technologies cannot accurately trace the source of anomalies in curved screen displays. Specifically, by establishing a deep correlation model between multi-source signals, it can achieve precise fault location from observed phenomena to specific physical areas and electrical channels, thereby providing a basis for implementing source-level targeted compensation. Summary of the Invention

[0005] The purpose of this invention is to provide a curved screen display method and apparatus to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a curved screen display method, the method comprising: Capture the original display signal stream corresponding to the target curved screen. The original display signal stream includes pixel signal sequences generated by multiple signal sources, display driving timing signal stream, and embedded sensor signal stream. The original display signal stream is synchronized, aligned, and standardized for encapsulation to generate a three-dimensional spatiotemporal data volume in a unified format; A dynamic anomaly perception map is constructed based on the three-dimensional spatiotemporal data volume. The nodes of the dynamic anomaly perception map are composed of visual attribute units of pixel signal sequences and logical state units of display driving timing signal streams. The edges of the dynamic anomaly perception map are composed of physical quantity change relationships of embedded sensor signal streams. The correlation between visual attribute unit nodes and logical state unit nodes in the dynamic anomaly perception map is mapped to a preset curved screen physical model, and the anomaly correlation strength between nodes is calculated. Based on the abnormal correlation strength, the abnormal source node is traced back in the dynamic abnormal perception map, and the screen physical area and signal driving channel corresponding to the abnormal source node are located. Based on the backtracking positioning results, a pixel signal compensation sequence and a driving parameter compensation sequence are generated for the physical area of ​​the screen and the signal driving channel.

[0007] Preferably, the construction of a dynamic anomaly perception map based on the three-dimensional spatiotemporal data volume includes: Extract pixel signal blocks within a continuous time window from the three-dimensional spatiotemporal data volume; Visual element encoding is performed on each pixel signal block to obtain shape encoding, edge direction encoding, and texture density encoding; The shape encoding, edge direction encoding, and texture density encoding are combined into a visual attribute encoding set; Based on the three-dimensional coordinates of the pixel signal block on the curved screen, a spatial location label is marked for each visual attribute encoding set; The set of visual attribute codes labeled with spatial location tags is registered as the visual attribute unit node of the dynamic anomaly perception map.

[0008] Preferably, the step of constructing a dynamic anomaly perception map based on the three-dimensional spatiotemporal data volume further includes: Extract display driving timing signal segments synchronized with the pixel signal sequence from the three-dimensional spatiotemporal data volume; Identify logic level transition events and steady-state holding intervals within the display driver timing signal segment; For each logic level transition event and steady-state holding interval, a logic state encoding is performed to generate a state encoding sequence; Based on the order of the state encoding sequence in the driving logic link, a logical state transition chain is constructed; Each logical state in the logical state transition chain is encoded and registered as a logical state unit node of the dynamic anomaly perception graph, and the state transition relationship is registered as a logical edge connecting the logical state unit nodes.

[0009] Preferably, the edges of the dynamic anomaly perception map are composed of the physical quantity changes of the embedded sensor signal stream, including: Obtain embedded sensor signal streams corresponding to the same time window from the three-dimensional spatiotemporal data volume; Analyze the physical quantity coupling relationship between different sensor signals in the embedded sensor signal stream; Establish multiple types of association edges between visual attribute unit nodes and logical state unit nodes. These multiple types of association edges include structural connection edges determined by spatial proximity, time dependency edges determined by time synchronization, and causal inference edges derived from the coupling relationship of physical quantities. Calculate the connection strength value for each type of associated edge, and inject the connection strength value as an attribute into the corresponding associated edge.

[0010] Preferably, the step of mapping the association relationship between visual attribute unit nodes and logical state unit nodes in the dynamic anomaly perception map to a preset curved screen physical model and calculating the anomaly association strength between nodes includes: The dynamic anomaly perception map is loaded into the curved screen physical model, which includes the screen stack-up structure, circuit routing layout and driver chip port mapping relationship. Under the constraints of the physical model of the curved screen, a multi-hop random walk from the logical state unit node to the visual attribute unit node is performed; Record the node paths traversed and the associated edges traversed in each random walk; Based on the number of physical layers that the node path passes through in the screen stack-up structure and circuit routing layout, the connection strength value of the associated edges that it passes through is attenuated. The abnormal association strength is obtained by accumulating the attenuated connection strength values ​​generated by all random walk paths between each pair of visual attribute unit nodes and logical state unit nodes.

[0011] Preferably, the step of tracing back the anomaly source node in the dynamic anomaly perception map based on the anomaly correlation strength, and locating the screen physical area and signal driving channel corresponding to the anomaly source node, includes: In the dynamic anomaly perception map, visual attribute unit nodes whose anomaly association strength exceeds a set threshold are marked as anomaly behavior nodes; Starting from the abnormal behavior node, perform graph diffusion calculation along the reverse of the associated edge to capture all logical state unit nodes affected by the abnormal behavior node, forming an abnormal diffusion subgraph. The state code of each logical state unit node in the abnormal diffusion subgraph is analyzed and compared with the preset normal logical state library to select logical state unit nodes with state deviation as candidate abnormal source nodes. Based on the circuit trace layout position of the candidate anomaly source node in the physical model of the curved screen, the corresponding signal drive channel number is derived. The coordinates of the affected screen physical region are determined based on the spatial location labels of visual attribute unit nodes that have strong correlation edges with the candidate anomaly source node.

[0012] Preferably, the step of selecting logical state unit nodes that deviate from their states as candidate anomaly source nodes includes: Obtain the state encoding sequence of the logical state unit node in the anomaly diffusion subgraph; Search for the reference state encoding sequence that has the highest matching degree with the state encoding sequence from the preset normal logic state library; Calculate the state deviation between the state coding sequence and the reference state coding sequence at each temporal position; Aggregate the state deviations of all time positions in the state coding sequence to generate an overall deviation score for the logical state unit node; Logical state unit nodes with an overall deviation score greater than zero are marked as logical state unit nodes with state deviation and added to the candidate anomaly source node set.

[0013] Preferably, the step of generating a pixel signal compensation sequence and a driving parameter compensation sequence for the screen physical area and signal driving channel based on the backtracking positioning result includes: Based on the physical area coordinates of the screen, extract the original pixel signal sequence of the target area from the three-dimensional spatiotemporal data volume; Based on the signal driving channel number, extract the original driving timing signal of the target channel from the three-dimensional spatiotemporal data volume; Based on the type of the anomaly source node, the corresponding signal compensation template and parameter compensation template are matched from the compensation strategy library; The original pixel signal sequence is convolved with the signal compensation template to generate an intermediate pixel signal compensation sequence. The original driving timing signal is superimposed with the parameter compensation template to generate an intermediate driving parameter compensation sequence. The intermediate pixel signal compensation sequence and the intermediate driving parameter compensation sequence are input into the physical model of the curved screen and the simulation is run to obtain the dynamic anomaly perception map after simulation. Calculate the attenuation of the abnormal correlation strength of the abnormal nodes in the simulated dynamic anomaly perception map; If the attenuation of the abnormal correlation strength does not reach the expected target, the weights of the signal compensation template and the parameter compensation template are adjusted, and the convolution operation, superposition operation and simulation steps are repeated until the termination condition is met, and the final pixel signal compensation sequence and driving parameter compensation sequence are output.

[0014] Preferably, the steps for constructing the preset curved screen physical model include: The physical structure parameters of the target curved screen are obtained, including the curvature radius of the screen substrate, the thickness and refractive index of each layer material, and the packaging position of the driver chip on the flexible circuit board. Based on the physical structure parameters, the curved surface of the screen substrate is reconstructed in a three-dimensional coordinate system, and based on the thickness and refractive index of each layered material, a screen stacked structure model including a polarizing layer, a liquid crystal layer, a color filter layer and an encapsulation layer is generated by stacking layers on the curved surface of the screen substrate. Import the circuit design file of the target curved screen, parse it to obtain the driver chip port definition, signal trace topology, and spatial crossing path of the trace in the screen stack-up structure model, and generate a circuit trace layout model. Establish a mapping table from each driver chip port in the circuit routing layout model to the corresponding pixel control unit in the screen stack-up structure model to complete the driver chip port mapping; The screen stack-up structure model, the circuit routing layout model, and the driver chip port mapping relationship are integrated and injected into a calculation engine based on material properties and circuit parameters to construct the physical model of the curved screen for simulating signal propagation and attenuation.

[0015] Preferably, the present invention further includes a curved screen display device, the curved screen display device including a processor and a memory, the memory and the processor being connected, the memory being used to store programs, instructions or code, and the processor being used to run the programs, instructions or code in the memory to implement the curved screen display method as described above.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By constructing a dynamic anomaly perception map that integrates pixel visual attributes, driving logic states, and sensor physical quantity changes, the barrier of isolated signal flow processing in traditional technologies is broken. The map uses nodes to represent state units of different dimensions and edges to dynamically define the actual coupling relationship between physical quantity changes and signal logic and visual performance, making the screen state a holistic, observable, and analyzable network. This structure can reveal in real time the intrinsic transmission chain of driving timing shifts induced by changes in physical conditions such as local deformation and temperature rise, ultimately manifesting as visual anomalies in specific pixel areas. This enables cross-domain correlation perception and early identification of complex display problems caused by multiple factors.

[0017] By mapping the relationships between nodes in the graph to a high-fidelity curved screen physical model for calculation, a quantified anomaly correlation strength can be obtained. This strength value is essentially a mathematical description of the probability and impact of a specific logical state anomaly causing a specific visual phenomenon, based on physical laws such as screen material properties, circuit layout, and signal propagation paths. Using this quantified strength for reverse tracing in the dynamic graph, it is possible to penetrate surface phenomena and directly trace back to the starting node of the anomaly-causing link. This process not only locates the physically abnormal area on the screen but also precisely pinpoints the driver chip that generates the error signal and even the specific output channel, achieving a fundamental tracing from "where is the problem with the image" to "which circuit component is causing the problem and why."

[0018] Based on precisely located screen physical areas and signal drive channel information, highly targeted compensation sequences can be generated. The pixel signal compensation sequence can reshape data for specific pixel areas damaged by source drive errors, while the drive parameter compensation sequence can directly calibrate the underlying parameters such as operating voltage and timing phase of the identified faulty channels. This dual compensation mechanism, addressing the root cause of the anomaly, avoids the secondary deviations that may arise from traditional overall or area-uniform compensation, ensuring a high degree of match between the correction action and the root cause of the fault. This improves the accuracy of display quality restoration and the overall energy efficiency of the system. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the working principle of the curved screen display method described in this invention. Figure 2 A flowchart for constructing visual attribute unit nodes of a dynamic anomaly perception map; Figure 3 A flowchart for constructing dynamic anomaly-aware graph edges; Figure 4 A graph showing the correlation analysis between the attenuation of abnormal correlation strength and the iterative adjustment parameters; Figure 5 This is a timing deviation analysis diagram for logic state unit nodes. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1This invention provides a curved screen display method, comprising: capturing an original display signal stream corresponding to a target curved screen, the original display signal stream including pixel signal sequences generated by multiple signal sources, a display driving timing signal stream, and an embedded sensor signal stream; performing synchronous alignment and standardized encapsulation on the captured original display signal stream to generate a unified format three-dimensional spatiotemporal data volume; constructing a dynamic anomaly perception map based on the three-dimensional spatiotemporal data volume, the nodes of which are composed of visual attribute units of pixel signal sequences and logical state units of display driving timing signal streams, and the edges of which are composed of physical quantity change relationships of embedded sensor signal streams; mapping the association relationships between visual attribute unit nodes and logical state unit nodes in the dynamic anomaly perception map to a preset curved screen physical model, and calculating the anomaly association strength between nodes under the constraints of this model; and, based on the calculated anomaly association strength, performing backtracking in the dynamic anomaly perception map to locate the anomaly source node and determine the screen physical area and signal driving channel corresponding to the anomaly source node. Based on the backtracking positioning results, a pixel signal compensation sequence and a driving parameter compensation sequence are generated for the positioned screen physical area and signal driving channel.

[0022] Example 1: See Figure 2 For each extracted pixel signal block, visual element encoding is performed to obtain shape encoding, edge direction encoding, and texture density encoding. These encodings are then combined into a visual attribute encoding set. Based on the 3D coordinates of the pixel signal block on the curved screen, a spatial location label is assigned to each visual attribute encoding set. The labeled visual attribute encoding sets are registered as visual attribute unit nodes in the dynamic anomaly perception map. Display driving timing signal segments synchronized with the pixel signal sequence are extracted from the 3D spatiotemporal data volume. Logic level transition events and steady-state intervals within these signal segments are identified. Logic state encoding is performed on each logic level transition event and steady-state interval to generate a state encoding sequence. A logic state transition chain is constructed based on the order of the state encoding sequences in the driving logic chain. Each logic state encoding in the logic state transition chain is registered as a logic state unit node in the dynamic anomaly perception map, and the state transition relationships are registered as logical edges connecting these nodes.

[0023] In the specific implementation, assuming the target curved screen is displaying a game scene containing fast-moving objects, the original display signal stream is captured and processed to generate a unified format of three-dimensional spatiotemporal data volume. A pixel signal block within a continuous time window is extracted from the three-dimensional spatiotemporal data volume. This pixel signal block contains 10 consecutive frames of screen data in the time dimension and corresponds to a 128x128 pixel rectangular area on the screen in the spatial dimension. Visual element encoding is performed on the pixel signal block. Shape encoding is achieved by identifying the contours of connected regions with similar colors within the pixel signal block and fitting them into geometric feature vectors. Edge direction encoding is obtained by performing multi-directional gradient operator convolution on the pixel signal block and calculating the principal gradient direction histogram. Texture density encoding is derived by calculating the contrast features of the gray-level co-occurrence matrix of the pixel signal block within the local window. The obtained shape encoding vector, edge direction encoding vector, and texture density encoding vector are concatenated and combined to form a visual attribute encoding set, which is a 768-dimensional feature vector. Based on the three-dimensional coordinates of pixel signal blocks on the curved screen, a curved UV mapping technique is used to convert the physical screen position of the pixel signal blocks into normalized three-dimensional coordinates, and a spatial location label is marked for each visual attribute encoding set. The visual attribute encoding sets marked with spatial location labels are registered as visual attribute unit nodes of the dynamic anomaly perception map, and each node stores the visual attribute encoding set and its corresponding three-dimensional coordinates.

[0024] In some embodiments, the processing of display driving timing signals is performed synchronously with the construction of visual attribute unit nodes. Display driving timing signal segments that are completely synchronized with the aforementioned pixel signal sequence are extracted from the three-dimensional spatiotemporal data volume. These signal segments include line scan signals, frame synchronization signals, and data enable signals. Logic level transition events and steady-state holding intervals are identified within the display driving timing signal segments. A logic level transition event is defined as a transient process in which the signal voltage jumps from a low-level threshold to a high-level threshold within 1 nanosecond. A steady-state holding interval is defined as the time period during which the signal voltage remains stable between two consecutive transition events. Logic state encoding is performed on each logic level transition event and steady-state holding interval. Transition events are encoded as event codes containing the transition edge type and timestamp, and steady-state holding intervals are encoded as state codes containing the level value and duration, thereby generating a state encoding sequence consisting of alternating event codes and state codes. Based on the order of the state encoding sequence in the driving logic chain, transition relationships between state codes are established, thus constructing a logic state transition chain, where each node represents a logic state. Each logical state code in the logical state transition chain is registered as a logical state unit node in the dynamic anomaly perception graph, and the transition relationship between the state codes is registered as a logical edge connecting the logical state unit nodes. The logical edge is directional, pointing from the prior state node to the successor state node.

[0025] Optionally, when constructing visual attribute unit nodes, the calculation of texture density encoding can be quantized using the following formula: in: This represents the calculated texture density encoded value. Represents the gray level in a pixel signal block. and It is a grayscale index. At a specified spatial offset Lower grayscale and The joint probability. It can be understood that the process of labeling the spatial location of the visual attribute encoding set takes into account the three-dimensional geometric deformation of the curved screen. Based on the physical area of ​​the screen corresponding to the pixel signal block, the two-dimensional pixel array coordinates are converted into three-dimensional spatial coordinates by querying the curvature mapping table of the curved screen's backplane. These three-dimensional coordinates reflect the actual position of the pixel on the curved surface.

[0026] Example 2: See Figure 3 The method involves extracting embedded sensor signal streams corresponding to the same time window from a 3D spatiotemporal data volume. It analyzes the physical quantity coupling relationships between different sensor signals within these embedded sensor signal streams. Various types of association edges are established between visual attribute unit nodes and logical state unit nodes. These include structural connection edges determined by spatial proximity, temporal dependency edges determined by temporal synchronization, and causal inference edges derived from physical quantity coupling relationships. A connection strength value is calculated for each type of association edge, and this value is injected as an attribute into the corresponding association edge. A dynamic anomaly perception map is loaded into a curved screen physical model, which includes the screen stack-up structure, circuit trace layout, and driver chip port mapping relationships. Under the constraints of the curved screen physical model, a multi-hop random walk is performed from logical state unit nodes to visual attribute unit nodes. The node paths traversed and the associated edges passed during each random walk are recorded. The connection strength values ​​of the traversed associated edges are attenuated based on the number of physical layers the node path passes through in the screen stack-up structure and circuit trace layout. The abnormal association strength is obtained by accumulating the attenuated connection strength values ​​generated by all random walk paths between each pair of visual attribute unit nodes and logical state unit nodes.

[0027] In practical implementation, within a continuous time window, in addition to pixel signal blocks and display driving timing signal segments, embedded sensor signal streams corresponding to the same time window are synchronously acquired from the three-dimensional spatiotemporal data volume. These embedded sensor signal streams contain readings from multiple sensors within the screen, such as signal sequences from flexible substrate deformation sensors, temperature sensors, and local capacitance sensors. The physical quantity coupling relationships between different sensor signals in the embedded sensor signal stream are analyzed. By calculating the cross-correlation function in the time domain between temperature sensor reading changes and capacitance sensor baseline drift, a positive correlation between temperature fluctuations and capacitance value changes is determined. Furthermore, by analyzing the correlation between the deformation sensor data change rate and specific driving timing stages, the potential impact of mechanical stress changes on the stability of the driving signal is inferred.

[0028] Multiple types of association edges are established between visual attribute unit nodes and logical state unit nodes in the dynamic anomaly perception map. Structural connection edges determined by spatial proximity calculate the Euclidean distance based on the spatial location labels of the visual attribute unit nodes; structural connection edges are established between node pairs whose distance is less than a set threshold. Temporal dependency edges determined by temporal synchronization are established between node pairs whose time difference is less than a synchronization tolerance by comparing the timestamps of the pixel signal blocks corresponding to the visual attribute unit nodes with the timestamps of the state encoding sequences corresponding to the logical state unit nodes. Causal inference edges derived from physical quantity coupling relationships transform the physical quantity influence paths obtained from embedded sensor signal flow analysis. For example, the coupling relationship between temperature and capacitance is mapped as a potential causal chain affecting the stability of the driving logic in a specific region, thereby establishing causal inference edges between relevant logical state unit nodes and downstream visual attribute unit nodes.

[0029] For each type of associated edge, a connection strength value is calculated and injected as an attribute into the corresponding associated edge. For structurally connected edges, the connection strength value is proportional to the reciprocal of the spatial distance between nodes. For temporally dependent edges, the connection strength value is determined by the tightness of time synchronization. For causal inference edges, the connection strength value is calculated based on the statistical significance and influence coefficient of the physical quantity coupling relationship. A final connection strength value is calculated. Here are some examples: in: This represents the final connection strength value of the injected associated edges. This represents the normalized spatial proximity score. This represents the normalized time synchronization score. This represents the normalized causal inference score. , , These are preset coefficients used to adjust the contribution weights of different types of edges, and .

[0030] In some embodiments, the dynamically constructed anomaly perception graph is loaded into a preset curved screen physical model. This model includes the screen's layered structure, circuit layout, and driver chip port mapping. Under the constraints of the preset curved screen physical model, a multi-hop random walk is performed from logical state unit nodes to visual attribute unit nodes. Each random walk starts from any logical state unit node in the graph and proceeds through multiple steps based on the transition probability determined by the direction and connection strength of the directed edges, until a visual attribute unit node is reached. The node paths traversed and the associated edges visited during each random walk are recorded. The node path information includes all logical state unit nodes and visual attribute unit nodes visited along the way, as well as the sequence of edges connecting them.

[0031] Based on the number of physical layers traversed by the node path in the screen's stack-up structure and circuit routing layout, the connection strength value of the associated edges is attenuated. In the preset curved screen physical model, signals attenuate when passing through different physical material layers; an attenuation factor is defined for each type of physical layer. For each associated edge traversed in the random walk path, the initial connection strength value of that edge is attenuated by multiplying the number of physical layers traversed by the edge in the physical model by the corresponding attenuation factor. The attenuated connection strength values ​​generated by all random walk paths between each pair of visual attribute unit nodes and logical state unit nodes are accumulated, and the attenuated connection strength values ​​of all paths connecting the same pair of nodes are summed to obtain the final abnormal association strength between that pair of nodes.

[0032] Optionally, the step size and number of repetitions for multi-hop random walks are preset. The maximum number of steps for a random walk is limited to prevent excessive computational complexity caused by over-walking. Random walks starting from each logical state unit node are repeated multiple times to fully explore all potential paths from that node to different visual attribute unit nodes, thereby obtaining a more stable and statistically significant cumulative value of anomaly correlation strength. It can be understood that the attenuation calculation of connection strength values ​​is tightly coupled with a preset curved screen physical model. This preset curved screen physical model not only provides a mapping of node spatial locations, but more importantly, its internally encapsulated material electromagnetic properties and signal transmission model provide a physical basis for quantifying the attenuation of signals or influences propagating in complex layered structures, making the connection strength attenuation calculated on the abstract graph physically realistic.

[0033] Example 3: In the dynamic anomaly perception graph, visual attribute unit nodes with anomaly association strength exceeding a set threshold are marked as anomaly manifestation nodes. Starting from the anomaly manifestation node, graph diffusion calculation is performed along the reverse of the association edge to capture all logical state unit nodes affected by the anomaly manifestation node, forming an anomaly diffusion subgraph. The state code of each logical state unit node in the anomaly diffusion subgraph is parsed and compared with a preset normal logical state library to filter out logical state unit nodes with state deviation as candidate anomaly source nodes. Based on the circuit trace layout position of the candidate anomaly source node in the curved screen physical model, the corresponding signal drive channel number is derived. Based on the spatial position labels of visual attribute unit nodes with strong association edges with the candidate anomaly source node, the coordinates of the affected screen physical area are determined. The state code sequence of the logical state unit nodes in the anomaly diffusion subgraph is obtained. The reference state code sequence with the highest matching degree with the state code sequence is found from the preset normal logical state library. The state deviation degree between the state code sequence and the reference state code sequence at each time position is calculated. The state deviation degree at all time positions in the state code sequence is aggregated to generate an overall deviation score for the logical state unit node. Logical state unit nodes with an overall deviation score greater than zero are marked as logical state unit nodes with state deviation and added to the candidate anomaly source node set.

[0034] In practical implementation, the dynamic anomaly perception graph has been constructed and the anomaly association strength between all relevant nodes has been calculated, with a predefined threshold value, such as 0.75. Within the dynamic anomaly perception graph, all visual attribute unit nodes are traversed, and those with anomaly association strength exceeding the set threshold are marked as anomalous behavior nodes. For example, a visual attribute unit node located in the lower right area of ​​the screen, displaying color block anomalies in its visual attribute encoding, has calculated anomaly association strengths of 0.82, 0.79, and 0.61 with multiple upstream logical state unit nodes. Since 0.82 and 0.79 exceed the set threshold of 0.75, this visual attribute unit node is marked as an anomalous behavior node. Starting from the anomalous behavior node, graph diffusion calculation is performed along the reverse of the association edges. The graph diffusion calculation uses a reverse random walk algorithm to capture all logical state unit nodes affected by the anomalous behavior node, forming an anomaly diffusion subgraph. The reverse walk starts from the abnormal behavior node, backtracks along the incoming edge to the upstream logical state unit node, and may continue to backtrack from these logical state unit nodes along the logical edge to the even upstream logical state unit node, until the preset diffusion depth is reached or the diffusion energy decays to below the threshold. All the nodes and edges traversed constitute the abnormal diffusion subgraph.

[0035] The state code of each logical state unit node in the anomaly propagation subgraph is analyzed and compared with a preset normal logical state library to filter out logical state unit nodes with state deviations as candidate anomaly source nodes. The preset normal logical state library stores the state code sequence patterns that the driving timing signals should have under standard operating conditions and various display modes. The actual state code sequences of the logical state unit nodes in the anomaly propagation subgraph are compared with the reference patterns in the normal logical state library. Based on the circuit trace layout position of the candidate anomaly source node in the curved screen physical model, the corresponding signal driving channel number is derived. The circuit trace layout model of the curved screen physical model accurately describes the connection relationship between each signal trace and the driver chip port. By querying the physical traces traversed by the driving signals represented by the logical state unit nodes in the circuit trace layout model, the specific driver chip port number can be mapped, thereby determining the signal driving channel number. Based on the spatial position labels of visual attribute unit nodes with strong correlation edges with the candidate anomaly source nodes, the coordinates of the affected screen physical area are determined. In the anomaly diffusion subgraph, visual attribute unit nodes with high connection strength values ​​between them and candidate anomaly source nodes are found. The spatial location labels of these visual attribute unit nodes are extracted, and these coordinate sets together delineate the physical area of ​​the screen affected by the candidate anomaly source node.

[0036] In some embodiments, the process of screening candidate anomaly source nodes includes detailed state deviation calculation. The state encoding sequence of the logic state unit nodes in the anomaly propagation subgraph is obtained. A reference state encoding sequence with the highest matching degree is searched from a preset normal logic state library. The state deviation between the state encoding sequence and the reference state encoding sequence at each timing position is calculated. The state deviation can be calculated by comparing the difference between the encoded values ​​at corresponding timing positions; for example, for the case where the encoding is a logic level value, the deviation can be defined as the absolute value of the difference between the actual level and the reference level. The state deviations at all timing positions in the state encoding sequence are aggregated to generate an overall deviation score for the logic state unit nodes. An overall deviation score is generated. The calculation example is as follows: in: This represents the overall deviation score of the logical state unit node. This indicates the length of the state-coded sequence (i.e., the total number of time positions). It is a time-series location index. Indicates the timing position The actual state coding value, Indicates the timing position The referenced state code value, It assigns a time position. The weighting coefficients are used to reflect the differences in the importance of states at different time stages. Logic state unit nodes with an overall deviation score greater than zero are marked as logic state unit nodes with state deviations and added to the candidate anomaly source node set.

[0037] Optionally, when constructing the anomaly diffusion subgraph, the energy attenuation model for graph diffusion calculation is related to the connection strength value and edge type of the associated edges. When diffusing backward along the associated edges, the diffusion energy is multiplied by an attenuation coefficient based on the edge type. Structural connection edges, time-dependent edges, and causal inference edges can have different attenuation coefficients. When the diffusion energy falls below a set cutoff threshold, the diffusion path terminates. This helps limit the scope of the anomaly's impact to directly and strongly related nodes, preventing excessive expansion of the subgraph. It can be understood that the construction of the preset normal logic state library relies on the long-term data collection and analysis of fault-free curved screens operating under various typical working scenarios. Statistical learning is used to extract the steady-state and transient characteristic patterns of various driving signals, forming a set of reference state encoding sequences, providing a benchmark for calculating state deviation.

[0038] Example 4: Based on the screen's physical region coordinates, extract the original pixel signal sequence of the target region from the 3D spatiotemporal data volume. Based on the signal driving channel number, extract the original driving timing signal of the target channel from the 3D spatiotemporal data volume. Based on the type of the anomaly source node, match the corresponding signal compensation template and parameter compensation template from the compensation strategy library. Perform convolution operation on the original pixel signal sequence and the signal compensation template to generate an intermediate pixel signal compensation sequence. Perform superposition operation on the original driving timing signal and the parameter compensation template to generate an intermediate driving parameter compensation sequence. Input the intermediate pixel signal compensation sequence and the intermediate driving parameter compensation sequence into the curved screen physical model to run the simulation and obtain the simulated dynamic anomaly perception map. Calculate the anomaly correlation strength attenuation of the anomaly manifestation nodes in the simulated dynamic anomaly perception map. If the anomaly correlation strength attenuation does not reach the expected target, adjust the weights of the signal compensation template and the parameter compensation template, and repeat the convolution operation, superposition operation, and simulation steps until the termination condition is met, and output the final pixel signal compensation sequence and driving parameter compensation sequence.

[0039] In specific implementation, assuming that through the steps of Example 3, an anomaly source node was identified by backtracking, this node corresponds to a physical area of ​​a dark spot in the upper left corner of the screen, with the screen physical area coordinates being a rectangle from (X1, Y1) to (X2, Y2), and the signal driving channel causing the anomaly was identified as data driving channel number 15. Based on the screen physical area coordinates (X1, Y1) to (X2, Y2), the original pixel signal sequence of the target area is extracted from the three-dimensional spatiotemporal data volume. This original pixel signal sequence contains data on the change of the RGB brightness value of each pixel in the rectangular area over time in the past few frames. Based on the signal driving channel number "15", the original driving timing signal of the target channel is extracted from the three-dimensional spatiotemporal data volume. This original driving timing signal is a sequence of voltage or current waveforms changing over time on the 15th output pin of the driver chip.

[0040] Based on the type of the anomaly source node, the corresponding signal compensation template and parameter compensation template are matched from the compensation strategy library. The type of the anomaly source node is identified in Example 3, for example, it is marked as "drive signal voltage drop". The compensation strategy library is a predefined lookup table that stores the mapping relationship between different anomaly types and recommended compensation schemes. Refer to Table 1, which shows a simplified compensation strategy library matching table: Table 1: Compensation Strategy Library Matching Table Exception types Compensation operation type Signal Compensation Template ID Parameter Compensation Template ID Scope of application Drive signal voltage drop Pixel gain enhancement T_pl_inc_002 P_v_comp_005 Data-driven channel Row scan timing jitter Pixel temporal interpolation T_ti_interp_007 P_t_delay_003 Scan drive channel Coupled noise interference Pixel filtering smoothing T_fl_median_010 P_n_supp_008 Data-driven channel Based on the matching results, select the signal compensation template ID "T_pl_inc_002" and the parameter compensation template ID "P_v_comp_005" for the "Drive Signal Voltage Sinking" type. The signal compensation template "T_pl_inc_002" is a two-dimensional filter kernel or pixel value transformation function, and the parameter compensation template "P_v_comp_005" is a voltage compensation waveform or a set of adjustment parameters.

[0041] The original pixel signal sequence is convolved with a signal compensation template to generate an intermediate pixel signal compensation sequence. The convolution operation is performed in the spatiotemporal domain. The signal compensation template acts on each pixel in the original pixel signal sequence and its spatiotemporal neighborhood. For example, for each pixel in the target region of each frame, its RGB value is multiplied by a gain-enhancing coefficient matrix to generate the brightness-enhanced intermediate pixel signal compensation sequence. The original driving timing signal is superimposed with a parameter compensation template to generate an intermediate driving parameter compensation sequence. The superposition operation can be waveform addition. For example, the parameter compensation template provides a positive voltage compensation pulse waveform, which is superimposed on the voltage waveform of the original driving timing signal at a specific timing point to raise its voltage level, thereby generating the intermediate driving parameter compensation sequence. The intermediate pixel signal compensation sequence and the intermediate driving parameter compensation sequence are input into a curved screen physical model for simulation to obtain the simulated dynamic anomaly perception map. The curved screen physical model simulates the propagation and final photoelectric conversion effect of the input compensated signal in the screen's stacked structure and circuitry. Based on the simulation output, it reconstructs the pixel signals and driving signals, and then reconstructs a simulated dynamic anomaly perception map according to the methods in Examples 1 and 2. The attenuation of the anomaly correlation strength of the anomaly-manifesting nodes in the simulated dynamic anomaly perception map is calculated.

[0042] If the attenuation of abnormal correlation strength does not reach the expected target, the weights of the signal compensation template and the parameter compensation template are adjusted, and the convolution operation, superposition operation, and simulation steps are repeated until the termination condition is met, outputting the final pixel signal compensation sequence and driving parameter compensation sequence. The expected target can be set to ΔI must be greater than a positive threshold. If ΔI is less than this threshold, it indicates that the compensation effect is insufficient. Adjusting the weights means modifying the gain coefficient in the signal compensation template or the voltage compensation amplitude in the parameter compensation template. A parameter compensation template voltage adjustment amplitude is used to calculate the (n+1)th iteration. The formula example is as follows: in: This represents the voltage adjustment magnitude of the parameter compensation template in the (n+1)th iteration. This represents the voltage adjustment magnitude of the parameter compensation template in the nth iteration. This represents a preset learning rate coefficient. This represents the preset target anomaly correlation strength attenuation amount. This represents the actual attenuation of the anomaly correlation strength calculated after the nth iteration of simulation. This adjustment process is repeated cyclically, and the termination condition can be that ΔI reaches the target, the number of iterations exceeds the upper limit, or the adjustment magnitude is less than the minimum change. When the termination condition is met, the intermediate pixel signal compensation sequence and the intermediate driving parameter compensation sequence generated in the last iteration are used as the final output results.

[0043] See Figure 4 In the iterative optimization process of anomaly compensation for curved screen displays, the dynamic changes and correlations of anomaly correlation strength attenuation, pixel gain coefficient, and voltage adjustment amplitude with the number of iterations are visually presented. Specifically, the graph uses the number of iterations as the horizontal axis to simultaneously show the evolution of three types of indicators: the anomaly correlation strength attenuation (dark blue line) increases stepwise with the advancement of iterations, especially showing a significant increase in the later stages of iterations (after 12 generations); the pixel gain coefficient (orange line) shows a continuous and gradual upward trend; and the voltage adjustment amplitude (yellow line) maintains a steady increase. The red dashed line in the graph marks the target value (0.8) for the anomaly correlation strength attenuation. As can be seen from the curve trend, with the increase of the number of iterations, the anomaly correlation strength attenuation gradually approaches this target value, confirming the effectiveness of the compensation template weight adjustment strategy—by iteratively optimizing the weights of the signal compensation template and the parameter compensation template, the attenuation effect of the anomaly correlation strength is continuously enhanced. Meanwhile, the coordinated change relationship of the three types of parameters also reflects the linkage mechanism of "pixel signal gain - driving parameter adjustment - abnormal correlation strength attenuation" in the compensation process: the continuous increase of pixel gain coefficient and voltage adjustment amplitude directly promotes the convergence of abnormal correlation strength attenuation to the target value. This dynamic process provides a quantitative basis and iterative optimization direction for the accurate compensation of abnormal curved screen display.

[0044] Example 5: Obtain the physical structure parameters of the target curved screen, including the curvature radius of the screen substrate, the thickness and refractive index of each layer material, and the packaging position of the driver chip on the flexible circuit board. Based on the physical structure parameters, the curved surface of the screen substrate is reconstructed in a three-dimensional coordinate system. Based on the thickness and refractive index of each layer material, a screen stacked structure model containing a polarizing layer, a liquid crystal layer, a color filter layer, and a packaging layer is generated by stacking layers on the curved surface of the screen substrate. The circuit design file of the target curved screen is imported, and the driver chip port definition, signal trace topology, and spatial crossing path of the traces in the screen stacked structure model are parsed to generate a circuit trace layout model. A mapping table is established from each driver chip port in the circuit trace layout model to the corresponding pixel control unit in the screen stacked structure model to complete the driver chip port mapping. The screen stacked structure model, the circuit trace layout model, and the driver chip port mapping relationship are integrated and injected into a calculation engine based on material properties and circuit parameters to construct a curved screen physical model for simulating signal propagation and attenuation.

[0045] In practical implementation, the physical structural parameters of the target curved screen are obtained. These parameters include the radius of curvature of the screen substrate, the thickness and refractive index of each layer material, and the packaging position of the driver chip on the flexible circuit board. For example, the radius of curvature of the screen substrate is 800 mm. The layer materials include a polarizing layer with a thickness of 0.2 mm and a refractive index of 1.5, a liquid crystal layer with a thickness of 0.005 mm, a color filter layer with a thickness of 0.015 mm and a refractive index of 1.52, and a packaging layer with a thickness of 0.1 mm and a refractive index of 1.45. The packaging position of the driver chip on the flexible circuit board is defined by the coordinates (15.2 mm, 3.8 mm) of the chip's center point relative to the lower left corner of the screen and a rotation angle of 0 degrees. Based on the physical structural parameters, the curved surface of the screen substrate is reconstructed in a three-dimensional coordinate system. With the screen center as the origin, and based on the radius of curvature of 800 mm and the screen size, a continuous curved surface model is generated through parametric surface equations. Based on the thickness and refractive index of each layer material, a screen stacked structure model containing a polarizing layer, a liquid crystal layer, a color filter layer and an encapsulation layer is generated by stacking layers one by one on the curved surface of the screen substrate. Each layer model not only includes the geometry, but also the optical and electrical properties of the materials.

[0046] Import the circuit design file for the target curved screen. This file is typically in Gerber or OBD++ format. Parse the circuit design file to obtain the driver chip port definitions, signal trace topology, and the spatial paths of the traces within the screen's multilayer structure model, generating a circuit trace layout model. The driver chip port definitions include the logic function, electrical characteristics, and physical pin coordinates of each port. The signal trace topology describes how clock lines, data lines, power lines, etc., connect the driver chip to the screen electrodes in the multilayer flexible circuit board. The spatial paths of the traces are mapped from the lamination structure information in the circuit design file to the corresponding three-dimensional spatial positions in the screen's multilayer structure model, forming a wire model with three-dimensional paths.

[0047] A mapping table is established to connect each driver chip port in the circuit routing layout model to the corresponding pixel control unit in the screen stack-up structure model, thus completing the driver chip port mapping. For example, the "data output port A15" of the driver chip is connected to the thin-film transistor gate of the pixel control unit in the 3rd row and 48th column of the screen stack-up structure model through a trace. This connection is recorded in the mapping table, and the table entries include the driver chip port ID, trace ID, the three-dimensional coordinates of the target pixel control unit, and the electrical interface type of the connection point.

[0048] This model integrates the screen's stacked structure model, circuit routing layout model, and driver chip port mapping relationships, and injects them into a computational engine based on material properties and circuit parameters to construct a physical model of the curved screen for simulating signal propagation and attenuation. The computational engine integrates a series of physical simulation models, such as the resistance-capacitance-inductance model for signal transmission in wires, the coupling model of an electric field passing through materials with different dielectric constants, and the orientation optical model of liquid crystal molecules under the influence of an electric field. It also includes a model for calculating the attenuation factor of signals due to dielectric loss in a specific stacked path. The formula example is as follows: in: The final attenuation factor representing the signal voltage or field strength. Indicates the frequency of the signal. This represents the speed of light in a vacuum. This indicates the i-th material layer that the signal path passes through. This represents the path length the signal travels through in the i-th material layer. Denotes the relative permittivity of the i-th material layer. This represents the loss tangent of the i-th material layer. This attenuation factor... It will be used to calculate the intensity change of the signal after it propagates from the driver chip port to the pixel control unit within the model.

[0049] See Figure 5 In the analysis of logic state unit nodes in the dynamic anomaly perception map, the timing deviation analysis diagram of logic state unit nodes presents the timing position-state value distribution characteristics of different nodes in stage 5. Specifically, the diagram uses timing position as the horizontal axis and state value as the vertical axis to distinguish the state performance of normal state nodes and abnormal nodes 1, 2, and 3: the state value of normal state nodes remains stable around 1.0, reflecting the steady-state characteristics of logic state units under normal operating conditions; while the state values ​​of abnormal nodes 1, 2, and 3 fluctuate significantly with timing position, with abnormal node 1 exhibiting the largest fluctuation (peak value above 0.85), abnormal node 3 having a generally low and frequent fluctuation, and abnormal node 2 falling in between. This quantitative representation of state deviation is based on the logic state encoding results of the display driver timing signal segment: by identifying logic level transition events and steady-state intervals, the state encoding sequence of each node is mapped to state values, and then compared using timing position as the dimension. The core value of this diagram lies in its intuitive presentation of the temporal deviation of logical state unit nodes, providing fundamental data support for subsequent calculations of anomaly correlation strength and tracing back to the source nodes of anomalies. In terms of parameters, the temporal positions cover 1-15 consecutive time windows, and the quantization range of state values ​​is consistent with the normalization standard of logical state encoding.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for displaying a curved screen, characterized in that, The method includes: Capture the original display signal stream corresponding to the target curved screen. The original display signal stream includes pixel signal sequences generated by multiple signal sources, display driving timing signal stream, and embedded sensor signal stream. The original display signal stream is synchronized, aligned, and standardized for encapsulation to generate a three-dimensional spatiotemporal data volume in a unified format; A dynamic anomaly perception map is constructed based on the three-dimensional spatiotemporal data volume. The nodes of the dynamic anomaly perception map are composed of visual attribute units of pixel signal sequences and logical state units of display driving timing signal streams. The edges of the dynamic anomaly perception map are composed of physical quantity change relationships of embedded sensor signal streams. The correlation between visual attribute unit nodes and logical state unit nodes in the dynamic anomaly perception map is mapped to a preset curved screen physical model, and the anomaly correlation strength between nodes is calculated. Based on the abnormal correlation strength, the abnormal source node is traced back in the dynamic abnormal perception map, and the screen physical area and signal driving channel corresponding to the abnormal source node are located. Based on the backtracking positioning results, a pixel signal compensation sequence and a driving parameter compensation sequence are generated for the physical area of ​​the screen and the signal driving channel.

2. The curved screen display method according to claim 1, characterized in that, The construction of a dynamic anomaly perception map based on the three-dimensional spatiotemporal data volume includes: Extract pixel signal blocks within a continuous time window from the three-dimensional spatiotemporal data volume; Visual element encoding is performed on each pixel signal block to obtain shape encoding, edge direction encoding, and texture density encoding; The shape encoding, edge direction encoding, and texture density encoding are combined into a visual attribute encoding set; Based on the three-dimensional coordinates of the pixel signal block on the curved screen, a spatial location label is marked for each visual attribute encoding set; The set of visual attribute codes labeled with spatial location tags is registered as the visual attribute unit node of the dynamic anomaly perception map.

3. The curved screen display method according to claim 1, characterized in that, The construction of the dynamic anomaly perception map based on the three-dimensional spatiotemporal data volume also includes: Extract display driving timing signal segments synchronized with the pixel signal sequence from the three-dimensional spatiotemporal data volume; Identify logic level transition events and steady-state holding intervals within the display driver timing signal segment; For each logic level transition event and steady-state holding interval, a logic state encoding is performed to generate a state encoding sequence; Based on the order of the state encoding sequence in the driving logic link, a logical state transition chain is constructed; Each logical state in the logical state transition chain is encoded and registered as a logical state unit node of the dynamic anomaly perception graph, and the state transition relationship is registered as a logical edge connecting the logical state unit nodes.

4. The curved screen display method according to any one of claims 2 and 3, characterized in that, The edges of the dynamic anomaly perception map are composed of the physical quantity changes in the embedded sensor signal flow, including: Obtain embedded sensor signal streams corresponding to the same time window from the three-dimensional spatiotemporal data volume; Analyze the physical quantity coupling relationship between different sensor signals in the embedded sensor signal stream; Establish multiple types of association edges between visual attribute unit nodes and logical state unit nodes. These multiple types of association edges include structural connection edges determined by spatial proximity, time dependency edges determined by time synchronization, and causal inference edges derived from the coupling relationship of physical quantities. Calculate the connection strength value for each type of associated edge, and inject the connection strength value as an attribute into the corresponding associated edge.

5. The curved screen display method according to claim 1, characterized in that, The step of mapping the association between visual attribute unit nodes and logical state unit nodes in the dynamic anomaly perception map to a preset curved screen physical model and calculating the anomaly association strength between nodes includes: The dynamic anomaly perception map is loaded into the curved screen physical model, which includes the screen stack-up structure, circuit routing layout and driver chip port mapping relationship. Under the constraints of the physical model of the curved screen, a multi-hop random walk from the logical state unit node to the visual attribute unit node is performed; Record the node paths traversed and the associated edges traversed in each random walk; Based on the number of physical layers that the node path passes through in the screen stack-up structure and circuit routing layout, the connection strength value of the associated edges that it passes through is attenuated. The abnormal association strength is obtained by accumulating the attenuated connection strength values ​​generated by all random walk paths between each pair of visual attribute unit nodes and logical state unit nodes.

6. The curved screen display method according to claim 1, characterized in that, The step of tracing back the anomaly source node in the dynamic anomaly perception map based on the anomaly correlation strength, and locating the screen physical area and signal driving channel corresponding to the anomaly source node, includes: In the dynamic anomaly perception map, visual attribute unit nodes whose anomaly association strength exceeds a set threshold are marked as anomaly behavior nodes; Starting from the abnormal behavior node, perform graph diffusion calculation along the reverse of the associated edge to capture all logical state unit nodes affected by the abnormal behavior node, forming an abnormal diffusion subgraph. The state code of each logical state unit node in the abnormal diffusion subgraph is analyzed and compared with the preset normal logical state library to select logical state unit nodes with state deviation as candidate abnormal source nodes. Based on the circuit trace layout position of the candidate anomaly source node in the physical model of the curved screen, the corresponding signal drive channel number is derived. The coordinates of the affected screen physical region are determined based on the spatial location labels of visual attribute unit nodes that have strong correlation edges with the candidate anomaly source node.

7. The curved screen display method according to claim 6, characterized in that, The logical state unit nodes that deviate from the selected state are used as candidate anomaly source nodes, including: Obtain the state encoding sequence of the logical state unit node in the anomaly diffusion subgraph; Search for the reference state encoding sequence that has the highest matching degree with the state encoding sequence from the preset normal logic state library; Calculate the state deviation between the state coding sequence and the reference state coding sequence at each temporal position; Aggregate the state deviations of all time positions in the state coding sequence to generate an overall deviation score for the logical state unit node; Logical state unit nodes with an overall deviation score greater than zero are marked as logical state unit nodes with state deviation and added to the candidate anomaly source node set.

8. The curved screen display method according to claim 1, characterized in that, The step of generating pixel signal compensation sequences and driving parameter compensation sequences for the screen physical area and signal driving channel based on the backtracking positioning results includes: Based on the physical area coordinates of the screen, extract the original pixel signal sequence of the target area from the three-dimensional spatiotemporal data volume; Based on the signal driving channel number, extract the original driving timing signal of the target channel from the three-dimensional spatiotemporal data volume; Based on the type of the anomaly source node, the corresponding signal compensation template and parameter compensation template are matched from the compensation strategy library; The original pixel signal sequence is convolved with the signal compensation template to generate an intermediate pixel signal compensation sequence. The original driving timing signal is superimposed with the parameter compensation template to generate an intermediate driving parameter compensation sequence. The intermediate pixel signal compensation sequence and the intermediate driving parameter compensation sequence are input into the physical model of the curved screen and the simulation is run to obtain the dynamic anomaly perception map after simulation. Calculate the attenuation of the abnormal correlation strength of the abnormal nodes in the simulated dynamic anomaly perception map; If the attenuation of the abnormal correlation strength does not reach the expected target, the weights of the signal compensation template and the parameter compensation template are adjusted, and the convolution operation, superposition operation and simulation steps are repeated until the termination condition is met, and the final pixel signal compensation sequence and driving parameter compensation sequence are output.

9. The curved screen display method according to claim 1, characterized in that, The steps for constructing the preset curved screen physical model include: The physical structure parameters of the target curved screen are obtained, including the curvature radius of the screen substrate, the thickness and refractive index of each layer material, and the packaging position of the driver chip on the flexible circuit board. Based on the physical structure parameters, the curved surface of the screen substrate is reconstructed in a three-dimensional coordinate system, and based on the thickness and refractive index of each layered material, a screen stacked structure model including a polarizing layer, a liquid crystal layer, a color filter layer and an encapsulation layer is generated by stacking layers on the curved surface of the screen substrate. Import the circuit design file of the target curved screen, parse it to obtain the driver chip port definition, signal trace topology, and spatial crossing path of the trace in the screen stack-up structure model, and generate a circuit trace layout model. Establish a mapping table from each driver chip port in the circuit routing layout model to the corresponding pixel control unit in the screen stack-up structure model to complete the driver chip port mapping; The screen stack-up structure model, the circuit routing layout model, and the driver chip port mapping relationship are integrated and injected into a calculation engine based on material properties and circuit parameters to construct the physical model of the curved screen for simulating signal propagation and attenuation.

10. A curved screen display device, characterized in that, The curved screen display device includes a processor and a memory, the memory and the processor are connected, the memory is used to store programs, instructions or code, and the processor is used to run the programs, instructions or code in the memory to implement the curved screen display method according to any one of claims 1 to 9.

Images