Augmented reality-based device preventive maintenance assistance work system

Abstract

The application discloses an augmented reality-based device preventive maintenance auxiliary work system, which comprises an augmented reality interaction module, a multi-source data acquisition module, an operation compliance verification module and a device state prediction module.The augmented reality interaction module is used for generating and superimposing virtual maintenance guide information on a real device image.The multi-source data acquisition module is used for collecting real-time operation state data of the device, operation data generated when a maintenance personnel performs an operation and environment data.The operation compliance verification module is in communication connection with the augmented reality interaction module and the multi-source data acquisition module, is used for verifying the compliance of the real-time collected operation data according to a preset maintenance procedure, and triggers an alarm when an exception is verified.The device state prediction module is used for receiving a verification result output by the operation compliance verification module and device historical performance data, performing calculation through a pre-trained prediction model, and outputting a comprehensive health degree score representing a future health state of the device after completing current maintenance.The application can realize efficient, accurate and scientific device preventive maintenance.

Description

Technical Field

[0001] This invention relates to the field of equipment maintenance, and more specifically to an augmented reality-based preventive maintenance assistance system for equipment. Background Technology

[0002] Preventive maintenance assistance is a set of auxiliary tasks conducted before equipment failure occurs to identify potential faults and ensure stable equipment operation. These tasks include operational guidance, compliance verification, status assessment, and strategy formulation. The core objectives are to improve maintenance efficiency, reduce failure rates, and optimize maintenance resource allocation. However, current assistance methods suffer from several pain points: operational compliance verification lacks precise, multi-dimensional quantitative methods, making it prone to misjudgment due to single parameter anomalies or deviations in non-critical parameters; equipment status prediction fails to effectively integrate historical status characteristics with current operational information, resulting in insufficient specificity of prediction model inputs and low accuracy in future health status assessments; maintenance plans often adopt standardized uniformity without dynamic adjustment based on the actual health level of the equipment, making it difficult to balance reliability and economy; virtual maintenance guidance information cannot be dynamically adapted to the perspective of maintenance personnel, limiting visibility and usability; and the selection of alternative maintenance strategies lacks comprehensive quantitative assessment of health gains, execution costs, and potential risks, resulting in insufficient scientific decision-making. These problems hinder the achievement of precision, intelligence, and efficiency in preventive maintenance. Therefore, this paper proposes an augmented reality-based preventive maintenance assistance system. Summary of the Invention

[0003] The present invention solves the above-mentioned technical problems through the following technical solutions, the present invention comprising: Augmented reality interaction module, used to generate and overlay virtual maintenance guidance information onto real device images; The multi-source data acquisition module is used to collect real-time operating status data of equipment, operation data generated by maintenance personnel, and environmental data. The operation compliance verification module communicates with the augmented reality interaction module and the multi-source data acquisition module. It is used to verify the compliance of the real-time collected operation data according to the preset maintenance procedures and trigger an alert when the verification is abnormal. The equipment status prediction module receives the verification results output by the operation compliance verification module and the historical performance data of the equipment. It calculates the data using a pre-trained prediction model and outputs a comprehensive health score S that characterizes the future health status of the equipment after the current maintenance is completed. The maintenance decision optimization module, which is connected to the equipment status prediction module and the augmented reality interaction module, is used to dynamically generate and recommend adjusted subsequent maintenance steps or alternative maintenance strategy schemes based on the comprehensive health score S.

[0004] Furthermore, the specific process by which the operational compliance verification module performs compliance verification is as follows: Obtain the standard operating requirements corresponding to the current maintenance step, which must include at least the standard parameter range and the standard operating sequence; Based on the operation data collected in real time by the multi-source data acquisition module, a current operation feature vector containing multiple dimensions is constructed, with each dimension corresponding to a specific operation parameter; Compare the current operation feature vector with the standard operation vector constructed according to the standard operation requirements; An abnormal operation is determined by calculating the overall difference between two vectors. The calculation process is as follows: first, the values ​​of each dimension of the vectors are normalized; then, weight factors are assigned to each operation parameter according to their importance; finally, the overall difference value is calculated using a weighted distance. The formula is as follows: ; Where D represents the overall difference degree, n is the number of vector dimensions, and wi is the weight factor corresponding to the i-th dimension parameter. and These are the normalized values ​​of the current operation feature vector and the standard operation vector in the i-th dimension, respectively; The calculated overall difference D is compared with the preset difference threshold D. th When comparing, when the overall difference is... If this occurs, the current operation is deemed abnormal.

[0005] Furthermore, after determining that an operation is abnormal, the operation compliance verification module specifically performs the following steps: Calculate the deviation values ​​of each dimension of the current operation feature vector from the corresponding dimensions of the standard operation vector, and analyze the contribution of each dimension deviation to the overall difference degree D, i.e., the contribution analysis results. Based on the contribution analysis results, identify one or more key dimension parameters that contribute the most to the anomaly. Based on key dimension parameters, the standard operating procedures and recommended corrective operating parameters corresponding to the parameters are retrieved from the pre-stored maintenance procedure knowledge base. Based on the search results, generate warning information that includes the identifiers of key dimension parameters, their standard value ranges, current measured values, and recommended correction operation parameters; The warning message is sent to the augmented reality interaction module, which then overlays the warning message in a highlighted and visualized manner onto the location of the operating component in the real device image that is associated with the key dimensional parameters.

[0006] Furthermore, the construction and operation process of the pre-trained prediction model in the equipment status prediction module includes: Model training phase: The model is trained using a historical dataset. Each sample in the historical dataset includes historical maintenance operation records, the temporal state characteristics of the equipment before the operation, and the actual changes in the equipment performance indicators over a period of time after the operation. Model inference phase: When processing the current maintenance task, the model receives the current operation characteristics and the recent historical state characteristics of the equipment as input; The model processes recent historical state features through an attention mechanism. The process involves: calculating the correlation between the current operational feature and the historical state features at each time step, obtained by calculating the cosine similarity between the two feature vectors; assigning different attention weights to each historical state feature based on this correlation, ultimately generating a weighted fusion of context-enhanced feature vectors. The formula for calculating the attention weights is as follows: ; in, Represents the historical state features at the j-th time step. The attention weights obtained The current operation feature vector is represented by sim, which represents the cosine similarity calculation function, and m is the total number of time steps for the historical state features. This represents the historical state feature vector at the k-th time step.

[0007] Furthermore, the calculation process of the comprehensive health score S output by the equipment status prediction module specifically includes: Obtain the future performance state prediction value Vpred, which is directly calculated by the prediction model based on the current input. This value represents the expected degradation of the key performance indicators of the equipment over a period of time in the future. Obtain the baseline health score HBase, which characterizes the basic health level of the equipment before this maintenance. This value is calculated based on the ratio of the equipment's historical fault-free operating time to the average level of similar equipment. Obtain the verification results output by the operation compliance verification module, and calculate an operation quality score Q that characterizes the overall standardization of this maintenance operation based on the results; The overall health score S is obtained by linearly weighting and fusing the above three quantities. The specific acquisition process is as follows: ; Wherein, α, β and γ are the fusion weight coefficients of the baseline health Hbase, the operational quality coefficient Q and the future performance state prediction value Vpred, respectively, and α+β+γ=1.

[0008] Furthermore, the process by which the maintenance decision optimization module dynamically generates adjustment suggestions based on the comprehensive health score S is as follows: Preset a first health threshold T1 and a second health threshold T2, where T1 < T2. The first health threshold T1 and the second health threshold T2 are obtained from the statistical analysis of the health score and the subsequent failure rate in the device's historical maintenance data. If the comprehensive health score S < T1, it is determined that the device's health status is poor. The maintenance decision optimization module generates an enhanced maintenance plan, which includes suggestions such as adding additional inspection items or adopting more stringent maintenance standards. If the comprehensive health score S > T2, it is determined that the device's health status is good. The maintenance decision optimization module generates an optimized maintenance plan, which includes suggestions such as simplifying or combining some non-critical maintenance steps. If T1 ≤ S ≤ T2, a maintenance plan consistent with the standard maintenance procedure is generated.

[0009] Furthermore, the method by which the maintenance decision optimization module generates and recommends alternative maintenance strategy plans includes: Based on the current device status and maintenance objectives, construct multiple candidate alternative maintenance strategies. For each alternative maintenance strategy Mi, simulate its execution process, and call the device status prediction module to estimate the health score that the device will reach after executing this strategy, denoted as the strategy estimated health score Si. At the same time, estimate the time and material costs Ci required to execute this strategy and its potential operation risk level Ri. After that, construct an evaluation function to obtain the utility evaluation value Ui of the strategy Mi, and conduct a comprehensive utility evaluation of each alternative strategy. This function is calculated by normalizing and weighted summing the health improvement, required cost, and potential risk brought by the strategy. Among them, the normalization reference value of the health improvement , the normalization reference value of the cost C ref , the normalization reference value of the risk R ref Are respectively determined based on the statistical maximum value of the data of all historical maintenance strategies. The calculation formula is: ; Among them, Ui represents the utility evaluation value of the strategy Mi, S is the current comprehensive health score, Is the weight coefficient of the corresponding item; Sort all alternative strategies according to the utility evaluation value Ui, and present the sorting result and key comparison information to the maintenance personnel through the augmented reality interaction module. Furthermore, the augmented reality interaction module dynamically adjusts the display mode of the virtual guidance information according to the real-time gaze focus of the maintenance personnel. The adjustment process includes: Real-time obtain the position data of the maintenance personnel's gaze focus in the augmented reality coordinate system provided by the multi-source data acquisition module. Calculate the spatial overlap between the display area of ​​the virtual guidance information to be displayed and the focus area in the coordinate system, denoted as ρ; When the spatial overlap ρ is lower than the preset visibility threshold ρ th When the visibility of the current guidance information is deemed insufficient, the display attributes of the information are adjusted accordingly. The degree of adjustment is negatively correlated with the spatial overlap, and the formula for calculating the degree of adjustment is as follows: ; Among them, A adj This indicates the degree of adjustment for the displayed attributes. K is a basic adjustment coefficient that is positively correlated with the importance of the information, and its value is preset according to the type of guidance information.

[0010] Compared with existing technologies, this invention has the following advantages: This augmented reality-based equipment preventive maintenance assistance system uses an augmented reality interaction module to achieve precise overlay and dynamic display adjustment of virtual maintenance guidance, improving operational visibility and convenience. An operational compliance verification module verifies operational compliance in real time and accurately warns of anomalies, reducing operational errors. Relying on the pre-trained model and attention mechanism of the equipment status prediction module, it accurately outputs a comprehensive equipment health score, helping to predict future health status. Then, the maintenance decision optimization module dynamically generates suitable maintenance plans based on the score and recommends alternative strategies that balance health gains, costs, and risks. Overall, this system achieves intelligent, precise, and efficient preventive maintenance, effectively reducing equipment failure rates and improving maintenance quality and efficiency. Attached Figure Description

[0011] Figure 1 This is a system block diagram of the present invention. Detailed Implementation

[0012] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0013] like Figure 1 As shown, this embodiment provides a technical solution: an augmented reality-based equipment preventive maintenance assistance system, comprising: Augmented reality interaction module, used to generate and overlay virtual maintenance guidance information onto real device images; The multi-source data acquisition module is used to collect real-time operating status data of equipment, operation data generated by maintenance personnel, and environmental data. The operation compliance verification module communicates with the augmented reality interaction module and the multi-source data acquisition module. It is used to verify the compliance of the real-time collected operation data according to the preset maintenance procedures and trigger an alert when the verification is abnormal. The equipment status prediction module receives the verification results output by the operation compliance verification module and the historical performance data of the equipment. It calculates the data using a pre-trained prediction model and outputs a comprehensive health score S that characterizes the future health status of the equipment after the current maintenance is completed. The maintenance decision optimization module, which is connected to the equipment status prediction module and the augmented reality interaction module, is used to dynamically generate and recommend adjusted subsequent maintenance steps or alternative maintenance strategy schemes based on the comprehensive health score S. The pre-trained prediction model is a hybrid neural network model based on Long Short-Term Memory (LSTM) and multi-head attention mechanism. Its specific structure, training process, and inference process are as follows: This hybrid model consists of an input layer, an LSTM temporal feature extraction module, a multi-head attention focusing module, and a fully connected output layer connected in series. The parameters and functions of each module are defined as follows: Input layer: Receives a normalized multi-dimensional feature vector. The feature dimensions correspond to the key parameters of equipment maintenance (in this scenario, it is 3-dimensional, corresponding to the upgrade package verification related features, upgrade voltage supply features, and upgrade ambient temperature features, respectively). The input vector format is [f1,f2,f3], where f1,f2,f3∈[0,1]. LSTM temporal feature extraction module: adopts a 2-layer bidirectional LSTM structure, with 64 hidden units in each layer, the activation function is tanh, and the forget gate threshold is set to 0.8.

[0014] This module is used to mine the temporal dependencies of historical state features. By processing the historical state features S1, S2, and S3 of the last three time steps (corresponding to the last three months) time step by time, it outputs a feature matrix containing long-term temporal correlation information (the dimension is 3×128, i.e. the number of time steps × the bidirectional LSTM output dimension). Multi-head attention focusing module: Four attention heads are set, each with a feature mapping dimension of 32. The LSTM output feature matrix is ​​transformed into three sets of feature vectors through linear transformation (used for association calculation, matching localization, and weighted fusion, respectively). Then, the association weights between the features of each historical time step and the current operation feature are calculated using the following logic: Step 1: Convert the current operation feature vector O cur The vector is mapped to a reference vector with the same dimension as the historical feature matrix through a linear transformation. Step 2: Calculate the similarity between the feature vector and the reference vector at each historical time step; Step 3: Perform softmax normalization on the similarity to obtain the association weights at each historical time step; Step 4: Weight the historical feature matrix according to the association weights and output the context-enhanced feature vector (3-dimensional, consistent with the input feature dimension, to ensure compatibility in subsequent calculations). Fully connected output layer: Contains two fully connected network layers. The first layer has 32 neurons and the activation function is ReLU. The second layer has 1 neuron and no activation function (for regression tasks). It outputs the original value of the performance degradation of the output device within a specified period (6 months in this scenario) (range [0, 0.5]).

[0015] Training dataset construction: Historical data from 500 firmware upgrades and maintenance sessions over the past 3 years were collected from the target device (smart router). Each data sample contains "historical time-series state sequence + current operational features + subsequent performance change labels": Historical time-series state sequence: Each sample contains historical state features at 3 time steps (consistent with the number of time steps in the inference stage), i.e. Each time step feature is a 3-dimensional normalized vector; Current operational characteristics: the normalized vector O of the 3D key parameters corresponding to a single maintenance. cur ; Performance Change Label: Signal transmission stability attenuation within 6 months after this maintenance (obtained through actual testing, original value range [0, 0.5]); Training parameter settings: Batch size is set to 32, epochs are 100, the Adam optimizer is used, the initial learning rate is 0.001, the learning rate decay strategy is to decay to 0.8 of the current value every 20 epochs, and the loss function is mean squared error (MSE), the formula is: (in This represents the actual performance degradation. (where N is the number of batch samples and is the model's predicted value). Model optimization and validation: The training set and validation set are divided into an 8:2 ratio. The loss of the validation set is monitored in real time during training. Training is stopped when the loss of the validation set does not decrease for 10 consecutive rounds. The mean absolute error (MAE) of the finally trained model on the validation set is ≤0.02, which meets the accuracy requirements for engineering applications.

[0016] Input the current operation feature vector The historical state characteristics of the last three time steps are S1, S2, and S3 (all normalized). Temporal correlation features are extracted by the LSTM module, and then the key historical information highly related to the current operation is focused by the multi-head attention module to generate a context-enhanced feature vector [0.5507,0.5,0.5]. The fully connected output layer outputs a raw value of 0.15 for the performance degradation over the next 6 months. After normalization, the predicted future performance value is obtained. This provides data support for subsequent comprehensive health scoring and maintenance decisions.

[0017] The specific process of the compliance verification module performing compliance verification is as follows: Obtain the standard operating requirements corresponding to the current maintenance step, which must include at least the standard parameter range and the standard operating sequence; Based on the operation data collected in real time by the multi-source data acquisition module, a current operation feature vector containing multiple dimensions is constructed, with each dimension corresponding to a specific operation parameter; Compare the current operation feature vector with the standard operation vector constructed according to the standard operation requirements; An abnormal operation is determined by calculating the overall difference between two vectors. The calculation process is as follows: first, the values ​​of each dimension of the vectors are normalized; then, weight factors are assigned to each operation parameter according to their importance; finally, the overall difference value is calculated using a weighted distance. The formula is as follows: ; Where D represents the overall difference degree, n is the number of vector dimensions, and wi is the weight factor corresponding to the i-th dimension parameter. and These are the normalized values ​​of the current operation feature vector and the standard operation vector in the i-th dimension, respectively; The calculated overall difference D is compared with the preset difference threshold D. th When comparing, when the overall difference is... If this occurs, the current operation is deemed abnormal; By clearly defining standard operating requirements (including parameter range and operation sequence), constructing multi-dimensional operation feature vectors, assigning weights according to parameter importance, and calculating the comprehensive difference degree after normalization, the overall deviation between the current operation and the standard can be accurately quantified. This avoids misjudgment of a single parameter or interference from deviations of non-critical parameters, ensuring the comprehensiveness, objectivity, and relevance of compliance verification, effectively improving the standardization and accuracy of maintenance operations, and reducing equipment failures caused by improper operation.

[0018] By clearly defining standard operating requirements (including parameter range and operation sequence), constructing multi-dimensional operation feature vectors, assigning weights according to parameter importance, and calculating the comprehensive difference degree after normalization, the overall deviation between the current operation and the standard can be accurately quantified. This avoids misjudgment of a single parameter or interference from deviations of non-critical parameters, ensuring the comprehensiveness, objectivity, and relevance of compliance verification, effectively improving the standardization and accuracy of maintenance operations, and reducing equipment failures caused by improper operation.

[0019] Taking the firmware upgrade and maintenance steps of a smart router as an example, the compliance of the current upgrade operation is verified. The core parameters and standard requirements are as follows: The vector dimension n=3 (3 key operational parameters); Standard operating requirements: Upgrade package verification value: standard range 95%-100%, weight w1=0.6 (core parameter, directly affecting firmware integrity); Upgrade voltage supply: standard range 5.0-5.2V, weight w2=0.3 (minor parameter, affecting upgrade stability); Upgraded ambient temperature: standard range 15-35℃, weight w3=0.1 (auxiliary parameter, affecting equipment heat dissipation); The preset difference threshold Dth = 0.2 (determined through historical maintenance data statistics; values ​​exceeding this threshold are considered abnormal). Normalization method: Min-max normalization is used, and the formula is as follows: (Eliminate the influence of different parameter dimensions); Normalized ideal value of standard operating vector (calculated at the midpoint of the standard range of each parameter): ; Example 1: The current operation is compliant; Real-time collected current operation data: Upgrade package verification value Upgrade voltage supply Upgrade ambient temperature ; Calculate the normalized value of the current operation's feature vector; ; Calculate the overall difference degree D; ; Substitute the data: ; The current operating data is adjusted as follows: upgrade package verification value 98%, upgrade voltage supply 5.1V, upgrade ambient temperature 25℃; Recalculate the normalized value: ; Overall Difference: ; Compared with the threshold: Since D≈0.0775<Dth=0.2, the current operation is deemed compliant.

[0020] Example 2: The current operation is non-compliant; Real-time collected current operation data: Upgrade package verification value (Below the standard lower limit), upgrade the voltage supply. Upgrade ambient temperature ; Calculate the normalized value of the current operation's feature vector; , (Same as Example 1) (Same as Example 1); Calculate the overall difference degree D; ; Compared with the threshold: Since D≈0.6971≥Dth=0.2, the current operation is determined to be abnormal, and the deviation of the core parameter (upgrade package verification value) is the main reason.

[0021] After determining that an operation is abnormal, the operation compliance verification module performs the following steps: Calculate the deviation values ​​of each dimension of the current operation feature vector from the corresponding dimensions of the standard operation vector, and analyze the contribution of each dimension deviation to the overall difference degree D, i.e., the contribution analysis results. Based on the contribution analysis results, identify one or more key dimension parameters that contribute the most to the anomaly. Based on key dimension parameters, the standard operating procedures and recommended corrective operating parameters corresponding to the parameters are retrieved from the pre-stored maintenance procedure knowledge base. Based on the search results, generate warning information that includes the identifiers of key dimension parameters, their standard value ranges, current measured values, and recommended correction operation parameters; The warning message is sent to the augmented reality interaction module, which controls it to overlay the warning message in a highlighted and visualized manner onto the location of the operating component in the real device image that is associated with the key dimension parameters; After identifying an operational anomaly, it can accurately pinpoint the key dimension parameters that contribute most to the anomaly, clarify the root cause of the deviation, and simultaneously retrieve the corresponding standard operating procedures and recommended correction parameters. It generates an alert containing key information and then overlays it onto the corresponding operating component through augmented reality visualization, allowing maintenance personnel to quickly understand the core of the problem, standard requirements, and correction direction. This avoids blind troubleshooting, significantly improves the efficiency and accuracy of anomaly rectification, and ensures compliance of maintenance operations.

[0022] For example, in scenarios involving firmware upgrades and maintenance of smart routers: The vector dimension n=3, and the weights w1=0.6 (upgrade package verification value, core parameter), w2=0.3 (upgrade voltage supply, secondary parameter), and w3=0.1 (upgrade ambient temperature, auxiliary parameter). Normalized value of standard operation vector ; Normalized value of the current operation feature vector in case of an anomaly; ; The overall difference degree D≈0.6971≥Dth=0.2, indicating an operational anomaly has been identified; Calculate the deviation and contribution of each dimension: Deviation values ​​for each dimension: ; ; ; Contribution calculation logic: The contribution of a certain dimension to the overall variance D = the squared weighted deviation of that dimension / the sum of the squared weighted deviations of all dimensions, the formula is: ; Substitute the data into the calculation: Dimension 1 Contribution: ; Dimension 2 Contribution: ; Dimension 3: Contribution ; Identify key dimension parameters: Based on the contribution analysis results, the upgrade package checksum (dimension 1) contributes 100% to the anomaly and is therefore identified as a key dimension parameter.

[0023] Search standards, specifications, and correction parameters: According to the pre-stored maintenance procedure knowledge base, the standard operating procedure for the upgrade package verification value is "the verification value must be between 95% and 100% to ensure that the firmware file is complete and undamaged". The recommended correction parameters are "re-download the corresponding router firmware upgrade package from the official channel and perform the verification operation until the verification value is ≥95%".

[0024] Generate alert message: The warning message reads: "Key parameters: upgrade package verification value; standard value range: 95%-100%; current measured value: 93%; recommended correction: re-download the official firmware upgrade package, complete the verification (verification value ≥ 95%), and then perform the upgrade."

[0025] AR Visual Alert: The warning message is sent to the augmented reality interaction module, which then displays the message as a red highlighted pop-up window overlaid on the location of the operation component corresponding to the "firmware upgrade interface" in the real router image, providing a direct and intuitive prompt to maintenance personnel.

[0026] The process of building and running the pre-trained prediction model in the equipment status prediction module includes: Model training phase: The model is trained using a historical dataset. Each sample in the historical dataset includes historical maintenance operation records, the temporal state characteristics of the equipment before the operation, and the actual changes in the equipment performance indicators over a period of time after the operation. Model inference phase: When processing the current maintenance task, the model receives the current operation characteristics and the recent historical state characteristics of the equipment as input; The model processes recent historical state features through an attention mechanism. The process involves: calculating the correlation between the current operational feature and the historical state features at each time step, obtained by calculating the cosine similarity between the two feature vectors; assigning different attention weights to each historical state feature based on this correlation, ultimately generating a weighted fusion of context-enhanced feature vectors. The formula for calculating the attention weights is as follows: ; Among them, among them, Represents the historical state features at the j-th time step. The attention weights obtained The current operation feature vector is represented by sim, which represents the cosine similarity calculation function, and m is the total number of time steps for the historical state features. This represents the historical state feature vector at the k-th time step; The prediction model is trained using historical datasets (including historical maintenance operation records, pre-operation temporal state features, and post-operation performance change data). During inference, the current operation features and recent historical state features are integrated. An attention mechanism is used to assign weights to each historical state based on cosine similarity. This focuses on key historical information that is highly relevant to the current operation, filters out irrelevant interference, and generates context-enhanced feature vectors. This improves the relevance and effectiveness of the model input, thereby increasing the accuracy of future performance state predictions. This provides reliable data support for subsequent comprehensive health scoring and ensures the accuracy of future health status assessments of the equipment.

[0027] For scenarios such as firmware upgrades and maintenance of smart routers, the core parameters are the same as described above, focusing on attention mechanism processing for model inference and context-enhanced feature vector generation. The parameters are disclosed as follows: The total number of time steps for historical state features is m=3 (all historical state features have been normalized after maintenance in the last 3 months): S1=[0.55,0.5,0.5] (1 month ago, corresponding to upgrade package verification, voltage supply, and ambient temperature characteristics), S2=[0.7,0.5,0.5] (2 months ago), S3=[0.4,0.5,0.5] (3 months ago); Current operation feature vector (After normalization): [0.6, 0.5, 0.5]; Cosine similarity formula: ( For vector dot product, (where the vector magnitude is...) Attention weight formula: .

[0028] Calculate the vector magnitude: ; ; ; ; Calculate the vector dot product: ; ; ; Calculate cosine similarity: ; ; ; Calculate attention weights: ; Summation of denominators: ; ; Generate context-enhanced feature vectors: Dimension 1 (related to upgrade package verification): 0.3356×0.55+0.3346×0.7+0.3298×0.4≈0.1846+0.2342+0.1319=0.5507; Dimension 2 (Voltage-related upgrade): 0.3356×0.5+0.3346×0.5+0.3298×0.5=0.5; Dimension 3 (related to ambient temperature): 0.3356×0.5+0.3346×0.5+0.3298×0.5=0.5; Final vector: [0.5507, 0.5, 0.5].

[0029] The calculation process for the comprehensive health score S output by the equipment status prediction module specifically includes: Obtain the future performance state prediction value Vpred, which is directly calculated by the prediction model based on the current input. This value represents the expected degradation of the key performance indicators of the equipment over a period of time in the future. Obtain the baseline health score HBase, which characterizes the basic health level of the equipment before this maintenance. This value is calculated based on the ratio of the equipment's historical fault-free operating time to the average level of similar equipment. Obtain the verification results output by the operation compliance verification module, and calculate an operation quality score Q that characterizes the overall standardization of this maintenance operation based on the results; The overall health score S is obtained by linearly weighting and fusing the above three quantities. The specific acquisition process is as follows: ; Where α, β and γ are the fusion weight coefficients of the baseline health Hbase, the operational quality coefficient Q and the future performance state prediction value Vpred, respectively, and α+β+γ=1; By linearly weighting and fusing the baseline health score HBase, the operation quality score Q, and the future performance status prediction value Vpred (fusion weight α+β+γ=1), this method comprehensively covers the basic status of the equipment before maintenance, the degree of standardization of maintenance operations, and the future performance degradation trend. It avoids the one-sidedness of single-dimensional evaluation and can accurately and comprehensively characterize the future health status of the equipment after maintenance. This provides an objective and reliable core quantitative basis for the maintenance decision optimization module, ensuring the scientific and targeted nature of subsequent maintenance plan adjustments.

[0030] In the scenario of firmware upgrade and maintenance for smart routers, the comprehensive health score S ranges from [0,1]. The higher the score, the better the future health status of the device. All parameters are fully disclosed and consistent with the previous text. HBase baseline health metrics: Router's historical mean time between failures (MTBF) before this maintenance was 180 days; average mean time between failures (MTBF) for similar devices was 150 days; HBase normalization interval is [0.5, 1.5]; normalization formula is […]. , This represents the original fault-free ratio of the equipment before this maintenance. Operational quality score Q-related: The operational compliance verification module outputs a comprehensive variance D=0.0775, and the historical maximum comprehensive variance Dmax=1 (statistics based on historical data from similar maintenance). Scoring formula. ; Future performance state prediction Vpred related: The pre-trained model receives context-enhanced feature vectors [0.5507, 0.5, 0.5], outputs the original attenuation of signal transmission stability over 6 months = 0.15, the original attenuation range [0, 0.5], and the normalization formula. ; Fusion weights: α=0.4, β=0.3, γ=0.3 (satisfying α+β+γ=1), comprehensive health score formula. .

[0031] Future performance state prediction value Vpred: The original decay value is derived by the model based on context-enhanced feature vector inference, reflecting the degree of future performance degradation. After normalization, a unified dimension is achieved. ; The baseline health level of HBase before this maintenance: First, calculate the original fault-free ratio, then normalize it to the [0,1] interval to reflect the basic health advantage of the equipment: ; Operation quality score Q: It is inversely transformed based on the operation compliance verification result. The smaller the difference, the higher the score, which represents the degree of operation specification: ; Comprehensive health score S: It is calculated by substituting into the weighted fusion formula, considering the combined influence of the three: S = 0.4×0.7 + 0.3×0.9225 - 0.3×0.3 = 0.288 + 0.28 + 0.27675 - 0.09 = 0.46675 ≈ 0.467 The process by which the maintenance decision optimization module dynamically generates adjustment suggestions based on the comprehensive health score S is as follows: Preset the first health threshold T1 and the second health threshold T2, and T1 < T2. The first health threshold T1 and the second health threshold T2 are obtained from the statistical analysis of the health scores and subsequent failure rates in the device's historical maintenance data; If the comprehensive health score S < T1, it is determined that the device's health status is poor, and the maintenance decision optimization module generates an enhanced maintenance plan, which includes suggestions such as adding additional inspection items or adopting more stringent maintenance standards; If the comprehensive health score S > T2, it is determined that the device's health status is good, and the maintenance decision optimization module generates an optimized maintenance plan, which includes suggestions such as simplifying or combining some non-critical maintenance steps; If T1 ≤ S ≤ T2, a maintenance plan consistent with the standard maintenance procedures is generated; By presetting the double health thresholds T1, T2 (T1 < T2) based on the statistical analysis of historical maintenance data, corresponding maintenance plans are dynamically generated according to the interval attribution of the comprehensive health score S, avoiding the one-sidedness of unified maintenance, achieving precise adaptation of "strengthening maintenance when the health status is poor, optimizing and simplifying when the status is good, and executing standards when the status is medium", not only ensuring the operational reliability of the device, but also reducing unnecessary maintenance costs and time losses, and improving the utilization efficiency of maintenance resources and the scientific nature of decision-making.

[0032] Intelligent router firmware upgrade and maintenance scenario; Comprehensive health score S: It is calculated above that S ≈ 0.467 (the value range is [0, 1], and the higher the score, the better the health status); Basis for setting double thresholds: Based on the historical maintenance data of the router in the past 3 years for 500 times, the correlation between the health score S and the subsequent failure rate in 6 months is statistically analyzed. When S < 0.4, the failure rate > 30%, and when S > 0.7, the failure rate < 5%. Therefore, the first health threshold T1 = 0.4 and the second health threshold T2 = 0.7 are set (satisfying T1 < T2); Maintenance plan rules: S < T1 → Enhancement maintenance plan, S > T2 → Optimization maintenance plan, T1 ≤ S ≤ T2 → Standard maintenance plan.

[0033] Threshold comparison with the current score: Given T1 = 0.4, T2 = 0.7, and the current comprehensive health score S ≈ 0.467, it is determined that T1 ≤ S ≤ T2 (0.4 ≤ 0.467 ≤ 0.7).

[0034] Generate a maintenance plan dynamically: Since T1 ≤ S ≤ T2, a plan consistent with the standard maintenance procedure is generated. The specific content is: "Perform subsequent operations according to the standard router firmware upgrade procedure, including: verifying the integrity of the upgrade package (verification value ≥ 95%); maintaining the upgrade voltage at 5.0 - 5.2V and the ambient temperature at 15 - 35°C; conducting 3 signal transmission stability tests after the upgrade (success rate ≥ 98%); 4. Record maintenance data in the device file and plan to perform a regular re-inspection after 6 months."

[0035] Supplement verification for two other interval scenarios: Scenario 1: If S = 0.35 (lower than T1 = 0.4, poor health status), generate an enhancement maintenance plan: Add the following on the basis of the standard procedure: Extra detection of the aging degree of the router power module (resistance value ≤ 5Ω); Clean the dust accumulated in the motherboard heat dissipation holes; Test the signal interference situation in adjacent frequency bands; 4. Shorten the re-inspection period to 3 months and continuously monitor the performance decay trend; Scenario 2: If S = 0.75 (higher than T2 = 0.7, good health status), generate an optimization maintenance plan: Simplify the standard procedure: Omit the repeated secondary verification of the upgrade package (only one verification value ≥ 95% is required); Combine the signal stability tests into 1 time (success rate ≥ 98%); Extend the re-inspection period to 9 months without additional detection items.

[0036] The method for the maintenance decision optimization module to generate and recommend alternative maintenance strategy plans includes: Construct multiple candidate alternative maintenance strategies based on the current device status and maintenance goals; For each alternative maintenance strategy Mi, simulate its execution process and call the device status prediction module to estimate the health score that the device will reach after executing this strategy, denoted as the strategy estimated health score Si. At the same time, estimate the time and material costs Ci required to execute this strategy and its potential operation risk level Ri; Then, an evaluation function is constructed to obtain the utility evaluation value Ui of strategy Mi. A comprehensive utility evaluation is performed on each candidate strategy. This function is obtained by normalizing and weighting the health gain, required cost, and potential risks brought by the strategy. Among them, the normalized reference value of the health gain is... The normalized reference value C for costs ref Normalized reference value R for risk ref The values ​​are determined based on the maximum statistical values ​​of all historical maintenance strategies, and the calculation formula is as follows: ; Where Ui represents the utility evaluation value of strategy Mi, and S is the current overall health score. These are the weighting coefficients for the corresponding items; All alternative strategies are ranked according to their utility evaluation value (Ui), and the ranking results and key comparison information are presented to maintenance personnel through an augmented reality interaction module.

[0037] By constructing multiple candidate maintenance strategies, simulating their execution, and comprehensively evaluating the health gains, required costs, and potential risks, the utility evaluation values ​​are calculated through normalization and weighted summation. Recommendations are then made in order of these evaluation values. This approach balances maintenance effectiveness, cost, and risk, avoids single-dimensional decision-making bias, provides maintenance personnel with optimal alternatives, enhances the scientific, economical, and targeted nature of maintenance strategy selection, and ensures efficient utilization of maintenance resources.

[0038] For example, in scenarios involving firmware upgrades and maintenance of smart routers: The current overall health score is S=0.467 (range [0,1]). The normalized reference value statistics are based on statistical analysis of router maintenance policy execution data from 500 operations over the past three years. Strategy-related definitions: Si is the strategy's estimated health score (output by the equipment status prediction module), Ci is the total cost of strategy execution (time cost + material cost, unit: yuan), and Ri is the potential operational risk level (level 1-5, with higher levels indicating greater risk).

[0039] Normalized reference value for health gain : Definition: ΔS = Si - S (the gain of health relative to the current state after policy execution); The maximum value of ΔS data from 500 historical maintenance strategies is [value missing]. (A certain enhanced maintenance strategy increased the health level from 0.3 to 0.65); Sure .

[0040] The normalized reference value for cost, Cref: Definition: Ci = Time cost (yuan) + Material cost (yuan) (Time cost is calculated at 100 yuan / hour, and material cost includes the cost of accessories and consumables). The Ci data from 500 historical maintenance strategies was analyzed, and the maximum value was Cmax = 300 yuan (for a certain full-item disassembly and testing strategy: 2 hours of labor + 100 yuan of consumables). Determine C ref =C max =300 yuan.

[0041] Normalized reference value R of risk ref : Definition: Ri is classified into levels based on operational complexity and probability of equipment failure (Level 1 is the lowest, Level 5 is the highest). The maximum value of Ri data from 500 historical maintenance strategies is R. max =Level 5 (A core component disassembly and replacement strategy carries the risk of motherboard damage). Determine R ref =R max =Level 5.

[0042] Construct candidate maintenance strategies: Strategy M1: Standard maintenance + additional motherboard thermal inspection (additional step of cleaning the heat dissipation holes); Strategy M2: Simplify upgrade package verification (only 1 verification) + extend the re-inspection cycle to 9 months; Strategy M3: Comprehensive enhanced testing (including power module, signal module, and heat dissipation system testing) + 3-month re-inspection.

[0043] Simulate execution and obtain the parameters of each strategy: Strategy M1: Estimated health score S1 = 0.58 (output of the device status prediction module); Cost C1 = 150 yuan (1 hour of labor + 50 yuan for heat dissipation cleaning consumables). Risk level R1=2 (heat dissipation cleaning is simple and there is no risk of damage). Strategy M2: Estimated health score S2 = 0.48 (output of equipment status prediction module); Cost C2 = 80 yuan (0.5 hours of labor + 30 yuan of basic consumables); Risk level R2 = Level 1 (simplified verification only, no additional operational risk). Strategy M3: Predicted health score S3 = 0.65 (output of equipment status prediction module); Cost C3 = 280 yuan (2 hours of labor + 80 yuan for all testing consumables); Risk level R3=4 (multi-module disassembly, with a risk of minor damage).

[0044] Calculate the utility evaluation value Ui for each strategy: Strategy M1: ; Strategy M2: ; Strategy M3: .

[0045] Strategy ranking and recommendation: The ranking is based on the utility evaluation value from highest to lowest: U1>U2>U3. Therefore, strategy M1 (standard maintenance + additional motherboard heat dissipation detection) is recommended first. The ranking results and key comparison information are presented to the maintenance personnel through the augmented reality interaction module (such as M1 having the best health gain and moderate cost, M2 having the lowest cost but limited gain, and M3 having the highest gain but high cost and risk).

[0046] The augmented reality interaction module dynamically adjusts the display of virtual guidance information based on the maintenance personnel's real-time gaze focus. This adjustment process includes: Real-time acquisition of the position data of the maintenance personnel's gaze focus in the augmented reality coordinate system, provided by the multi-source data acquisition module; Calculate the spatial overlap between the display area of ​​the virtual guidance information to be displayed and the focus area in the coordinate system, denoted as ρ; When the spatial overlap ρ is lower than the preset visibility threshold ρ th When the visibility of the current guidance information is deemed insufficient, the display attributes of the information are adjusted accordingly. The degree of adjustment is negatively correlated with the spatial overlap, and the formula for calculating the degree of adjustment is as follows: ; Among them, A adj This indicates the degree of adjustment for the displayed attributes. K is a basic adjustment coefficient that is positively correlated with the importance of the information, and its value is preset according to the type of guidance information. By acquiring the position of the maintenance personnel's gaze focus in real time, calculating the spatial overlap between the virtual guidance information and the gaze focus, and dynamically adjusting the display attributes based on the overlap (the degree of adjustment is negatively correlated with the overlap and the importance of related information), it is possible to avoid the guidance information being obscured or having insufficient visibility, ensuring that maintenance personnel can clearly obtain key guidance from different operating perspectives, improving the adaptability and effectiveness of virtual guidance, reducing the probability of operational errors, and ensuring smooth and efficient maintenance operations.

[0047] Continuing with the scenario of firmware upgrades and maintenance for smart routers, focusing on the display adjustment logic of the augmented reality interaction module, the core parameters are disclosed as follows: Augmented Reality Coordinate System Setting: Establish a three-dimensional coordinate system with the center of the router as the origin (X-axis horizontal left and right, Y-axis vertical up and down, Z-axis front and back depth), unit: cm; Virtual guidance information: The current key guidance to be displayed is "Upgrade package verification value must be ≥95%". The display area is a rectangular pop-up window (the coordinate range of the 3D projection onto the XY plane is: X∈[5,8], Y∈[3,6], that is, the horizontal width of the pop-up window is 3cm and the vertical height is 3cm). Key parameter definitions: Spatial overlap ρ (value range [0,1], ρ=1 indicates complete overlap, ρ=0 indicates no overlap), visibility threshold ρth=0.3 (below this value is considered insufficient visibility), basic adjustment coefficient K=0.8 (because this guideline is a key operational requirement, the K value is positively correlated with the importance of the information and is preset). Adjustment rules: Adjust the attribute to pop-up brightness (initial brightness 500 cd / m²). 2 Adjustment range 300-800 cd / m 2 The adjustment degree Aadj is positively correlated with the brightness increase. The brightness adjustment formula is: (300 represents the maximum brightness increase). Display the adjustment formula: .

[0048] Obtain gaze focus position data: The multi-source data acquisition module captures the eye movements of maintenance personnel in real time and outputs the XY plane projection coordinates of the gaze focus in the AR coordinate system: P=(6.5,7.2) (i.e. X=6.5cm, Y=7.2cm).

[0049] Calculate the spatial overlap ρ: The XY plane projection range of the virtual guidance pop-up is: X∈[5,8], Y∈[3,6]. The Y coordinate of the gaze focus is 7.2cm, which is outside the Y axis range of the pop-up, while the X coordinate is 6.5cm, which is within the X axis range of the pop-up. Therefore, it is determined that there is no spatial overlap. Spatial overlap calculation (based on projected area ratio): Overlap area = 0 cm 2 The projected area of ​​the pop-up window is 3 × 3 = 9 cm² 2 Therefore .

[0050] Determine visibility and calculate the degree of adjustment: Comparing ρ and ρth: ρ=0<ρth=0.3, it is determined that the visibility of the current guidance information is insufficient; Substitute into the adjustment formula to calculate: .

[0051] Perform display attribute adjustments: Calculate the target brightness using the brightness adjustment formula: ; The augmented reality interaction module will reduce the brightness of the pop-up window displaying "Upgrade package verification value must be ≥95%" from 500 cd / m². 2 Increased to 740 cd / m 2 At the same time, the pop-up window is slightly moved upward (the Y-axis coordinate is adjusted to [4.2,7.2]) to ensure that the overlap between the gaze focus P=(6.5,7.2) and the pop-up window is increased to ρ=0.4, which meets the visibility requirements.

[0052] Additional reverse scenario verification: If the focus coordinates are P=(6.2,4.5) (completely within the pop-up window), then ρ=1. The pop-up window maintains an initial brightness of 500 cd / m². 2 No additional adjustments are needed, avoiding excessive display that could interfere with operation.

[0053] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An augmented reality-based equipment preventive maintenance auxiliary operation system, characterized in that, include: Augmented reality interaction module, used to generate and overlay virtual maintenance guidance information onto real device images; The multi-source data acquisition module is used to collect real-time operating status data of equipment, operation data generated by maintenance personnel, and environmental data. The operation compliance verification module communicates with the augmented reality interaction module and the multi-source data acquisition module. It is used to verify the compliance of the real-time collected operation data according to the preset maintenance procedures and trigger an alert when the verification is abnormal. The equipment status prediction module receives the verification results output by the operation compliance verification module and the historical performance data of the equipment. It calculates the data using a pre-trained prediction model and outputs a comprehensive health score that characterizes the future health status of the equipment after the current maintenance is completed. The maintenance decision optimization module, which is connected to the equipment status prediction module and the augmented reality interaction module, is used to dynamically generate and recommend adjusted subsequent maintenance steps or alternative maintenance strategy schemes based on the comprehensive health score.

2. The augmented reality-based equipment preventive maintenance assistance system according to claim 1, characterized in that: The specific process of the compliance verification module performing compliance verification is as follows: Obtain the standard operating requirements corresponding to the current maintenance step, which must include at least the standard parameter range and the standard operating sequence; Based on the operation data collected in real time by the multi-source data acquisition module, a current operation feature vector containing multiple dimensions is constructed, with each dimension corresponding to a specific operation parameter; Compare the current operation feature vector with the standard operation vector constructed according to the standard operation requirements; The abnormality of an operation is determined by calculating the comprehensive difference between two vectors. The calculation process of the comprehensive difference is as follows: first, the values ​​of each dimension of the vector are normalized; then, weight factors are assigned to each operation parameter according to their importance; finally, the comprehensive difference D is obtained by calculating the weighted distance. The calculated overall difference D is compared with the preset difference threshold D. th When comparing, when the overall difference is... If this occurs, the current operation is deemed abnormal.

3. The augmented reality-based equipment preventive maintenance assistance system according to claim 2, characterized in that: After determining that an operation is abnormal, the operation compliance verification module performs the following steps: Calculate the deviation values ​​of each dimension of the current operation feature vector from the corresponding dimensions of the standard operation vector, and analyze the contribution of each dimension deviation to the overall difference degree D, i.e., the contribution analysis results. Based on the contribution analysis results, identify one or more key dimension parameters that contribute the most to the anomaly. Based on key dimension parameters, the standard operating procedures and recommended corrective operating parameters corresponding to the parameters are retrieved from the pre-stored maintenance procedure knowledge base. Based on the search results, generate warning information that includes the identifiers of key dimension parameters, their standard value ranges, current measured values, and recommended correction operation parameters; The warning message is sent to the augmented reality interaction module, which then overlays the warning message in a highlighted and visualized manner onto the location of the operating component in the real device image that is associated with the key dimensional parameters.

4. The augmented reality-based equipment preventive maintenance assistance system according to claim 1, characterized in that: The process of building and running the pre-trained prediction model in the equipment status prediction module includes: Model training phase: The model is trained using a historical dataset. Each sample in the historical dataset includes historical maintenance operation records, the temporal state characteristics of the equipment before the operation, and the actual changes in the equipment performance indicators within a preset time after the operation. Model inference phase: When processing the current maintenance task, the model receives the current operation characteristics and the historical state characteristics of the equipment within the recent preset time period as input; The model processes the historical state features within a recent preset time duration through an attention mechanism. The processing process is as follows: Calculate the correlation degree between the current operation feature and the historical state features at each time step, and the correlation degree is obtained by calculating the cosine similarity of the two feature vectors; Assign different attention weights to each historical state feature according to this correlation degree, and finally generate a weighted and fused context-enhanced feature vector.

5. The augmented reality-based equipment preventive maintenance assistance system according to claim 4, characterized in that: The calculation process of the comprehensive health score S output by the device state prediction module specifically includes: Obtain the future performance state prediction value Vpred directly calculated by the prediction model based on the current input, and this value represents the expected decay amount of the key performance indicators of the device within a future preset time duration; Obtain the baseline health score Hbase representing the basic health level of the device before this maintenance, and this value is calculated based on the ratio of the device's historical fault-free operation time to the average level of similar devices; Obtain the verification result output by the operation compliance verification module, and calculate an operation quality score Q representing the overall standard degree of this maintenance operation based on this result; The comprehensive health score S is obtained through linear weighted fusion calculation of the future performance state prediction value Vpred, the baseline health score Hbase before this maintenance, and the operation quality score Q.

6. The augmented reality-based equipment preventive maintenance assistance system according to claim 5, characterized in that: The process of the maintenance decision optimization module dynamically generating adjustment suggestions according to the comprehensive health score S is as follows: Preset the first health threshold T1 and the second health threshold T2, and T1 < T2. The first health threshold T1 and the second health threshold T2 are obtained from the statistical analysis of the health score and the subsequent failure rate in the device's historical maintenance data; If the comprehensive health score S < T1, it is determined that the device health state is poor, and the maintenance decision optimization module generates an enhanced maintenance plan, and the plan includes suggestions such as adding additional inspection items or adopting more stringent maintenance standards; If the comprehensive health score S > T2, it is determined that the device health state is good, and the maintenance decision optimization module generates an optimized maintenance plan, and the plan includes suggestions such as simplifying or combining some non-critical maintenance steps; If T1 ≤ S ≤ T2, a maintenance plan consistent with the standard maintenance procedure is generated.

7. The augmented reality-based equipment preventive maintenance assistance system according to claim 6, characterized in that: The method for the maintenance decision optimization module to generate and recommend alternative maintenance strategy plans includes: Based on the current device state and maintenance objectives, construct multiple candidate alternative maintenance strategies; For each alternative maintenance strategy Mi, simulate its execution process, and call the device state prediction module to estimate the health score that the device will reach after executing this strategy, denoted as the strategy estimated health score Si. At the same time, estimate the time and material costs Ci required to execute this strategy and its potential operation risk level Ri; After that, construct an evaluation function to obtain the utility evaluation value Ui of the strategy Mi to comprehensively evaluate each alternative strategy. This function is obtained by normalizing and weighted summing the health improvement, required cost, and potential risk brought by the strategy; Sort all alternative strategies according to the utility evaluation value Ui, and present the sorting result and key comparison information to the maintenance personnel through the augmented reality interaction module.

8. The augmented reality-based equipment preventive maintenance assistance system according to claim 7, characterized in that: The augmented reality interaction module dynamically adjusts the display of virtual guidance information based on the maintenance personnel's real-time gaze focus. This adjustment process includes: Real-time acquisition of the position data of the maintenance personnel's gaze focus in the augmented reality coordinate system, provided by the multi-source data acquisition module; Calculate the spatial overlap ρ between the display area of ​​the virtual guidance information to be displayed and the focus area in the coordinate system; When the spatial overlap ρ is lower than the preset visibility threshold ρ th When the visibility of the current guidance information is deemed insufficient, the display attributes of the information are adjusted accordingly, and the degree of adjustment is negatively correlated with the degree of spatial overlap.

Images