Power construction safety risk point level-to-level management method and system based on intelligent glasses

Abstract

The invention discloses an electric power construction safety risk point level-to-level management method and system based on intelligent glasses, and the method comprises the steps: collecting the multi-modal data of an electric power construction site through the intelligent glasses, and the multi-modal data comprise a construction scene image, a construction worker operation behavior video, environment data and position data; performing feature extraction on the construction scene image and the construction personnel operation behavior video based on a pre-trained deep learning model to obtain risk target features and behavior mode features; extracting abnormal environment features from the environment data; generating a feature vector set by combining the risk target feature, the behavior pattern feature, the abnormal environment feature and the position data; calculating a comprehensive risk evaluation value based on the feature vector set; mapping the comprehensive risk evaluation value based on a preset risk level threshold interval to obtain a corresponding risk level; and generating corresponding grading early warning instructions based on different risk grades.

Description

Technical Field

[0001] This invention relates to the field of power construction safety technology, and mainly to a method and system for hierarchical management of power construction safety risk points based on smart glasses. Background Technology

[0002] Electrical construction is a high-risk operation, often requiring personnel to install, inspect, and maintain equipment in high-altitude, high-voltage, and complex environments. Traditional safety management relies primarily on manual inspections, adherence to safety regulations, and post-event analysis, which has the following significant limitations: 1. Insufficient real-time monitoring: Safety officers cannot monitor all work sites and personnel around the clock, leading to delays in risk detection and handling; 2. High reliance on subjectivity: Risk identification heavily depends on the personal experience and sense of responsibility of supervisors, making it prone to omissions due to fatigue or negligence; 3. Limited information dimensions: Traditional methods struggle to simultaneously acquire and comprehensively analyze multi-dimensional risk information, including behavioral, environmental, and locational data, hindering a comprehensive and accurate risk assessment; 4. Outdated early warning methods: Traditional early warning methods such as warning signs and verbal reminders are inefficient and fail to ensure effective communication with workers in noisy or emergency construction environments.

[0003] Currently, smart glasses, due to their first-person perspective and hands-free operation, are beginning to be used in fields such as industrial inspection and remote assistance. Some attempts have been made to use them for safety monitoring, such as identifying whether safety helmets are being worn. However, most existing applications are single-function, typically limited to passive identification and simple alarms for single risk factors (such as the wearing of personal protective equipment), failing to form a proactive and systematic safety management solution integrating multimodal data perception, real-time risk fusion analysis, dynamic hierarchical assessment, and precise differentiated early warning.

[0004] Therefore, there is an urgent need for a risk point classification management method that can deeply integrate the hardware capabilities of smart glasses with advanced data analysis models to achieve real-time, automatic, accurate perception and hierarchical control of complex risk factors in the power construction process. Summary of the Invention

[0005] To address the problems existing in the prior art, this invention proposes a hierarchical management method and system for safety risk points in power construction based on smart glasses.

[0006] The technical solution of the present invention is as follows: On one hand, this invention proposes a hierarchical management method for safety risk points in power construction based on smart glasses. The smart glasses include an image acquisition unit, an environmental sensor unit, a positioning unit, a display unit, an audio unit, and a tactile feedback unit. The method is characterized by comprising: Smart glasses are used to collect multimodal data from power construction sites, including images of the construction scene, videos of construction workers' operational behavior, environmental data, and location data. Based on a pre-trained deep learning model, features are extracted from construction scene images and videos of construction workers' operational behaviors to obtain risk target features and behavioral pattern features; abnormal environmental features are extracted from environmental data. A feature vector set is generated by combining risk target characteristics, behavioral pattern characteristics, abnormal environment characteristics, and location data; a comprehensive risk assessment value is calculated based on the feature vector set. The comprehensive risk assessment value is mapped based on a preset risk level threshold range to obtain the corresponding risk level; corresponding graded early warning instructions are generated based on different risk levels.

[0007] Preferably, it also includes preprocessing of multimodal data, including data cleaning, alignment and normalization; Data cleaning is performed on the multimodal data to remove noisy and invalid data segments, and missing values ​​are imputed and repaired. The cleaned multimodal data is aligned based on a unified time base to form a time-synchronized multimodal data sequence; The aligned multimodal data sequences are normalized to eliminate the dimensional differences between different sensors.

[0008] Preferably, the image acquisition unit is used to acquire images of the construction scene and videos of the construction workers' operating behavior. The images of the construction scene are used to capture the equipment status, tool placement, and the integrity of safety signs. The videos of the construction workers' operating behavior are used to record the construction workers' operating actions and work processes. Environmental data is obtained by using environmental sensor units to monitor the temperature, humidity, concentration of harmful gases, noise level and light intensity of the construction site in real time. The location coordinates of construction personnel and key equipment are obtained using the positioning unit.

[0009] Preferably, the risk target characteristics include wearing a safety helmet, using insulated tools, and high-altitude operation procedures; The behavioral pattern characteristics include the amplitude of operation, the frequency of action, and the sequence of behavior. The abnormal environment characteristics are environmental data that exceed a preset safety threshold.

[0010] Preferably, the comprehensive risk assessment value is calculated based on the feature vector set, and the specific steps are as follows: Based on preset identification rules, the risk source category of each feature vector in the feature vector set is identified. The risk source category includes personnel behavior risk, equipment status risk, environmental anomaly risk, and regional location risk. Based on the feature vector set corresponding to each risk source category and the historical accident features in the risk database, the probability of risk occurrence for the current risk source category is calculated. Predict the severity of accident consequences for each risk source category based on pre-defined consequence assessment rules; The duration of risk source categories is extracted from the feature vector set to obtain the risk exposure time; The probability of risk occurrence, the severity of accident consequences, and the exposure time of risk are calculated using a weighted algorithm to obtain an independent risk value for each risk source. Aggregate the independent risk values ​​of all risk source categories and calculate the comprehensive risk assessment value using a preset comprehensive assessment function.

[0011] Preferably, the comprehensive risk assessment value is mapped based on a preset risk level threshold range, specifically as follows: When the comprehensive risk assessment value falls within the threshold range of the first risk level, it is mapped to a low risk level; When the comprehensive risk assessment value falls within the second risk level threshold range, it is mapped to a general risk level; When the comprehensive risk assessment value falls within the threshold range of the third risk level, it is mapped to a medium risk level; When the comprehensive risk assessment value falls within the threshold range of the fourth risk level, it is mapped to a high risk level.

[0012] Preferably, corresponding tiered early warning instructions are generated based on different risk levels, specifically as follows: When the risk level is low, a level one warning command is generated, and a soft text prompt icon is displayed at the edge of the field of vision through the display unit of the smart glasses, and a gentle voice reminder is played through the audio unit; When the risk level is general risk, a level 2 warning instruction is generated, a conspicuous text warning message is displayed in the center of the field of vision through the display unit, a level 1 voice warning is played in a loop through the audio unit, and a level 1 vibration is generated through the tactile feedback unit. When the risk level is medium risk, a level three warning instruction is generated. The display unit flashes a bright color to show a detailed risk description and handling suggestions. The audio unit plays a level two voice warning and the tactile feedback unit generates a level two vibration. When the risk level is high, a level four warning instruction is generated. The red emergency alarm information and emergency evacuation instructions are displayed in full-screen mode through the display unit, a level three voice warning is played through the audio unit, and a level three vibration is generated through the haptic feedback unit.

[0013] Preferably, the generation process of the graded early warning instruction also includes adaptively adjusting the display position and prompt intensity of the early warning content by combining the real-time working status and location information of the construction personnel.

[0014] Preferably, it also includes recording and uploading the entire process data of the current risk identification event to the remote security management platform, which is used to update the consequence assessment rules and risk database, optimize the calculation process of the comprehensive risk assessment value, and generate a security management report.

[0015] On the other hand, the present invention also provides a power construction safety risk point classification management system based on smart glasses, wherein the smart glasses include an image acquisition unit, an environmental sensor unit, a positioning unit, a display unit, an audio unit, and a tactile feedback unit, and the system includes: The data acquisition module uses smart glasses to collect multimodal data from power construction sites, including images of construction scenes, videos of construction workers' operational behavior, environmental data, and location data. The feature extraction module extracts features from construction scene images and videos of construction workers' operational behaviors based on a pre-trained deep learning model, obtaining risk target features and behavioral pattern features; it also extracts abnormal environmental features from environmental data. The risk assessment module generates a feature vector set by combining risk target characteristics, behavioral pattern characteristics, abnormal environment characteristics, and location data; and calculates a comprehensive risk assessment value based on the feature vector set. The instruction generation module maps the comprehensive risk assessment value based on a preset risk level threshold range to obtain the corresponding risk level; and generates corresponding graded early warning instructions based on different risk levels.

[0016] The present invention has the following beneficial effects: 1. This invention provides a hierarchical management method and system for safety risk points in power construction based on smart glasses. By automatically and continuously collecting multimodal data on site through smart glasses and using backend models for real-time analysis, it completely changes the lagging supervision mode that relies on manual inspections, realizes 24 / 7 uninterrupted automatic monitoring of safety risks, and greatly improves the timeliness of risk detection. 2. This invention provides a hierarchical management method and system for safety risk points in power construction based on smart glasses. It adopts a deep learning-based approach and integrates multiple factors to replace the traditional model that mainly relies on human subjective experience. It realizes quantitative calculation of risk probability, consequences, exposure time and other dimensions, making the risk assessment results more scientific, objective and accurate, and effectively reducing false alarms and missed alarms. 3. This invention provides a hierarchical management method and system for power construction safety risk points based on smart glasses. It integrates four major risk sources—personnel behavior, equipment status, environmental anomalies, and spatial location—into a unified framework for collaborative analysis and comprehensive judgment. This multi-dimensional and all-round perception capability can discover complex related risks that cannot be identified by single-dimensional monitoring, and achieve a global grasp of the safety status of the construction site. 4. This invention provides a method and system for hierarchical management of safety risks in power construction based on smart glasses. It adopts a differentiated early warning strategy that is precisely matched with the risk level, and provides multi-sensory prompts through the display, audio and tactile modules of the smart glasses. This early warning method not only frees up the hands of construction workers, but also ensures that the early warning information can be perceived and understood in the most effective way under different environments (such as noisy sites) and different levels of urgency, thereby guiding personnel to react quickly and correctly and improving the actual effect of early warning. Attached Figure Description

[0017] Figure 1 This is a detailed flowchart of an embodiment of the present invention. Detailed Implementation

[0018] 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.

[0019] It should be understood that the step numbers used in the text are for ease of description only and are not intended to limit the order in which the steps are performed.

[0020] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0021] The terms “comprising” and “including” indicate the presence of the described feature, whole, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.

[0022] The term “and / or” refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes these combinations.

[0023] Example 1: See Figure 1 This invention provides a hierarchical management method for safety risk points in power construction based on smart glasses, the method comprising: The smart glasses, acting as a front-end sensing node, integrate various sensors, each with its own function, including an image acquisition unit, an environmental sensor unit, a positioning unit, a display unit, an audio unit, and a tactile feedback unit. The image acquisition unit provides continuous and objective recording by mimicking human vision, and is used to capture static targets (equipment, tools) and dynamic processes (human behavior) in space. Environmental sensor units extend human sensory capabilities, enabling the perception of environmental factors (such as harmful gases and noise decibels) that cannot be accurately quantified or continuously perceived by the human ear or body, thus digitizing these invisible or easily overlooked risks. The positioning unit assigns spatial context to all data; based on positioning technologies such as GPS, BeiDou, or UWB, it determines the exact location where the risk occurs; S1. Use smart glasses to collect multimodal data from power construction sites, including construction scene images, videos of construction personnel's operational behavior, environmental data, and location data; S11. Use the image acquisition unit to acquire images of the construction scene and videos of the construction personnel's operation behavior. The images of the construction scene are used to capture the status of equipment, the placement of tools, and the integrity of safety signs. The videos of the construction personnel's operation behavior are used to record the construction personnel's operation actions and work processes. Environmental data is obtained by using environmental sensor units to monitor the temperature, humidity, concentration of harmful gases, noise level and light intensity of the construction site in real time. The location coordinates of construction personnel and key equipment are obtained using a positioning unit; Multimodal data were initially correlated based on timestamps to form multimodal data with spatiotemporal labels. This followed the principle of "spatiotemporal consistency," ensuring that images, environmental, and location data collected at the same time described the same scene, thus laying the foundation for subsequent multi-feature fusion analysis. S12 also includes preprocessing of multimodal data, including data cleaning, alignment and normalization; Data cleaning is performed on the multimodal data to remove noisy and invalid data segments, and missing values ​​are imputed and repaired. The cleaned multimodal data is aligned based on a unified time base to form a time-synchronized multimodal data sequence; The aligned multimodal data sequence is normalized to eliminate the dimensional differences between different sensors; S2. Based on a pre-trained deep learning model, feature extraction is performed on construction scene images and videos of construction workers' operational behaviors to obtain risk target features and behavioral pattern features; abnormal environmental features are extracted from environmental data. The risk target characteristics include wearing safety helmets, using insulated tools, and high-altitude operation procedures; The behavioral pattern characteristics include the amplitude of operation, the frequency of action, and the sequence of behavior. The abnormal environmental characteristics are environmental data that exceed a preset safety threshold. In this embodiment, for images or videos, a pre-trained convolutional neural network (e.g., CNN) is used for feature extraction; the principle is to automatically learn and extract features from pixels from low to high and from concrete to abstract through multi-layer convolution and pooling operations. S3. Generate a feature vector set by combining risk target characteristics, behavioral pattern characteristics, abnormal environment characteristics, and location data; S4. Calculate the comprehensive risk assessment value based on the feature vector set; S41. Based on preset identification rules, identify the risk source category of each feature vector in the feature vector set. The risk source category includes personnel behavior risk, equipment status risk, environmental anomaly risk, and regional location risk. In this embodiment, rule templates for each risk source category are constructed, wherein: Calculate the similarity between the feature vector set and the rule template of each risk source category, and select the corresponding risk source categories with similarity greater than a preset similarity threshold as candidate categories; If there is only one candidate category, then the current candidate category will be used as the risk source category; If the candidate categories are not unique, the candidate category with the highest similarity will be taken as the risk source category. S42. Based on the feature vector set corresponding to each risk source category and the historical accident features in the risk database, calculate the probability of risk occurrence for the current risk source category; In this embodiment, historical accident cases with the same category as the current risk source are retrieved from the risk database, and the feature vector set of the cases is obtained. The risk database stores historical accident data, including accident type, feature vector, occurrence scenario and actual occurrence probability label. Calculate the similarity score between the feature vector of the current risk source and each historical accident feature, wherein the calculation process uses one or more of the following methods in combination: including cosine similarity, Euclidean distance and pre-trained neural network model (such as Siamese network). Based on the similarity score, the probability of occurrence of the current risk source is obtained by one of the following methods, including the weighted average method: the similarity score of each historical accident feature is weighted and averaged with its actual occurrence probability label, with the weight being the similarity score itself, to obtain the probability of occurrence of the current risk source. Probabilistic model method: Use classification models (such as logistic regression, random forest or neural network) as input and output the probability of occurrence of risk for the current risk source category; Threshold method: Set a similarity threshold. When the similarity score exceeds the threshold, the average probability of occurrence of historical accidents is used as the probability of occurrence of the current risk source category. S43. Based on preset consequence assessment rules, predict the severity of accident consequences for each risk source category, specifically the severity of personal injury, equipment damage, and power outage range that may be caused if an accident of the current risk source category occurs. In this embodiment, the consequence assessment rules define multiple consequence assessment dimensions; the dimensions include at least: the personal safety dimension for assessing the level of injury that may be caused to construction workers, the equipment asset dimension for assessing the level of damage and economic loss that may be caused to power equipment, and the power grid operation dimension for assessing the scope and level of impact that may be caused to the reliability and stability of the power supply system. The feature vector of the risk source is input into the assessment sub-model corresponding to each consequence assessment dimension; the assessment sub-model outputs the severity level of the risk source in the corresponding dimension based on a preset rule base; the severity level includes at least four levels: "minor", "moderate", "serious" and "very serious"; The rule base contains expert rules composed of IF-THEN logic, for example: IF Risk source category is personnel behavior risk AND Operation height characteristic >10 meters THEN The severity level of the personal safety dimension is "severe"; The severity score and assessment weight of each consequence dimension are weighted and calculated to obtain the severity of the accident consequences, expressed by the formula: ; In the formula, Indicates the first The severity of the consequences of an accident at each risk source; Indicates the first The first risk source is in the The evaluation weights for each consequence dimension; Indicates the first The first risk source is in the The severity rating of each consequence dimension; S44. Statistically extract the persistence time of risk source categories from the feature vector set to obtain the risk exposure time; the longer the risk exposure time, the higher the overall risk. In another embodiment, the expected exposure time of construction workers in the feature vector set can also be statistically analyzed to obtain the risk exposure time; S45. Calculate the probability of risk occurrence, the severity of accident consequences, and the risk exposure time using a weighted algorithm to obtain an independent risk value for each risk source. The independent risk value quantifies the overall size of the risk, allowing risks of different natures to be compared with each other. The weighted algorithm may include linear weighting, fuzzy logic, etc. In this embodiment, a simple linear weighting is used, expressed by the formula: ; In the formula, Indicates the first The independent risk value of each risk source; Indicates the first The probability of risk occurring from each risk source; Indicates the first The risk exposure time of each risk source; This represents the pre-defined probability weight of the risk occurring; Indicates the pre-defined weights for the severity of accident consequences; This indicates the pre-defined risk exposure time weight; S46. Aggregate the independent risk values ​​of all risk source categories, and calculate the comprehensive risk assessment value using a preset comprehensive assessment function, expressed as the formula: ; In the formula, This represents the overall risk assessment value; Indicates the number of risk source categories; In step S4, based on the updated data issued by the remote security management platform, the weight parameters and consequence assessment rules used in the calculation process will be adaptively optimized periodically using the full-process data of risk identification events. S5. Map the comprehensive risk assessment value based on the preset risk level threshold range to obtain the corresponding risk level; When the comprehensive risk assessment value falls within the first risk level threshold range, it is mapped to a low risk level, indicating that there are minor safety hazards that require daily attention from construction personnel. When the comprehensive risk assessment value falls into the second risk level threshold range, it is mapped to a general risk level, indicating that there is a risk within an acceptable range, and construction personnel need to pay attention to it. When the comprehensive risk assessment value falls into the third risk level threshold range, it is mapped to a medium risk level, indicating that there are significant safety hazards, which should immediately alert the construction personnel and prepare to take countermeasures. When the comprehensive risk assessment value falls into the fourth risk level threshold range, it is mapped to a high risk level, indicating that there is a serious and imminent security threat and that emergency response actions must be taken immediately. S6. Generate corresponding graded early warning instructions based on different risk levels; Based on the risk level obtained from the mapping, a pre-set early warning rule base is invoked to generate a corresponding graded early warning instruction. The early warning rule base defines the combination of early warning content, early warning intensity, and feedback method corresponding to each risk level. When the risk level is low, a level one warning command is generated, and a soft text prompt icon is displayed at the edge of the field of vision through the display unit of the smart glasses, and a gentle voice reminder is played through the audio unit; When the risk level is general risk, a level 2 warning instruction is generated, a conspicuous text warning message is displayed in the center of the field of vision through the display unit, a level 1 voice warning is played in a loop through the audio unit, and a level 1 vibration is generated through the tactile feedback unit. When the risk level is medium risk, a level three warning instruction is generated. The display unit flashes a bright color to show a detailed risk description and handling suggestions. The audio unit plays a level two voice warning and the tactile feedback unit generates a level two vibration. When the risk level is high, a level four warning instruction is generated. The red emergency alarm information and emergency evacuation instructions are displayed in full-screen mode through the display unit. The level three voice warning is played through the audio unit and the level three vibration is generated through the haptic feedback unit. S7. The generation process of the graded early warning instruction also includes adaptively adjusting the display position and prompt intensity of the early warning content by combining the real-time working status and location information of the construction personnel, so as to ensure that the early warning information can be perceived by the construction personnel in a timely and effective manner. S8 also includes recording and uploading the entire process data of the current risk identification event to the remote security management platform, which is used to update the consequence assessment rules and risk database, optimize the calculation process of the comprehensive risk assessment value, and generate a security management report; The full-process data includes at least multimodal data, feature vector sets, comprehensive risk assessment values, risk levels, warning instruction content, and execution status. Accurate timestamps and spatial location markers are added to the data throughout the process, and the data is encapsulated according to a preset format to generate data packets. The encapsulated data packets are then uploaded to a remote security management platform via a wireless communication network. Data compression and encryption transmission technologies are used during the upload process to ensure data security and transmission efficiency. The remote security management platform receives and parses data packets, stores the parsed full-process data in the security event database, and triggers the data processing flow. At the same time, based on preset cycles or triggering conditions, a safety management report containing risk statistics, trend analysis, and rectification suggestions is automatically generated and pushed to relevant management personnel. The remote security management platform also provides data query and backtracking functions, supporting the retrieval and analysis of historical risk events by multiple dimensions such as time, location, and risk level.

[0024] Example 2: This embodiment provides a graded management system for safety risk points in power construction based on smart glasses. The smart glasses include an image acquisition unit, an environmental sensor unit, a positioning unit, a display unit, an audio unit, and a tactile feedback unit. The system includes: The data acquisition module uses smart glasses to collect multimodal data from power construction sites, including images of construction scenes, videos of construction workers' operational behavior, environmental data, and location data. The feature extraction module extracts features from construction scene images and videos of construction workers' operational behaviors based on a pre-trained deep learning model, obtaining risk target features and behavioral pattern features; it also extracts abnormal environmental features from environmental data. The risk assessment module generates a feature vector set by combining risk target characteristics, behavioral pattern characteristics, abnormal environment characteristics, and location data; and calculates a comprehensive risk assessment value based on the feature vector set. The instruction generation module maps the comprehensive risk assessment value based on a preset risk level threshold range to obtain the corresponding risk level; and generates corresponding graded early warning instructions based on different risk levels.

[0025] In this application embodiment, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent the existence of A alone, A and B simultaneously, or B alone. A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of singular or plural items. For example, at least one of a, b, and c can represent: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, and c can be single or multiple.

[0026] Those skilled in the art will recognize that the units and algorithm steps described in the embodiments disclosed herein can be implemented using electronic hardware, computer software, or a combination of electronic hardware and software. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0027] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0028] In the several embodiments provided in this application, any function, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0029] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A power construction safety risk point grading management method based on smart glasses, the smart glasses comprising an image acquisition unit, an environmental sensor unit, a positioning unit, a display unit, an audio unit, and a tactile feedback unit, characterized in that, The method comprises: Collecting multi-modal data of the power construction site using smart glasses, including construction scene images, construction personnel operation behavior videos, environmental data and location data; Extracting features from the construction scene images and the construction personnel operation behavior videos based on a pre-trained deep learning model to obtain risk target features and behavior pattern features; extracting abnormal environmental features from the environmental data; Generating a feature vector set by combining the risk target features, the behavior pattern features, the abnormal environmental features and the location data; calculating a comprehensive risk evaluation value based on the feature vector set; Mapping the comprehensive risk evaluation value based on a pre-set risk level threshold interval to obtain a corresponding risk level; generating a corresponding graded warning instruction based on different risk levels. 2.The smart glasses-based power construction safety risk point grading management method according to claim 1, characterized in that, It also includes preprocessing the multi-modal data, including data cleaning, alignment and normalization; Cleaning the multi-modal data to eliminate noise data and invalid data segments, and supplementing and repairing missing values; Aligning the cleaned multi-modal data based on a unified time reference to form a time-synchronized multi-modal data sequence; Normalizing the aligned multi-modal data sequence to eliminate dimension differences of different sensors. 3.The smart glasses-based power construction safety risk point grading management method of claim 1, wherein, Obtain construction scene images and construction personnel operation behavior videos using an image acquisition unit, wherein the construction scene images are used to capture equipment status, tool placement and safety sign integrity; the construction personnel operation behavior videos are used to record the operation actions and work processes of the construction personnel; Real-time monitoring of temperature and humidity, harmful gas concentration, noise level and light intensity of the construction site using an environmental sensor unit to obtain environmental data; Obtaining the location coordinates of the construction personnel and key equipment using a positioning unit. 4.The smart glasses-based power construction safety risk point grading management method of claim 1, wherein, The risk target features include safety helmet wearing, insulation tool use and aerial work specification; The behavior pattern features include operation amplitude, action frequency and behavior sequence; The abnormal environmental features are environmental data exceeding the pre-set safety threshold. 5.The smart glasses-based power construction safety risk point grading management method according to claim 1, characterized in that, The specific steps for calculating the comprehensive risk evaluation value based on the feature vector set are: Identifying the risk source category of each feature vector in the feature vector set based on a pre-set identification rule, wherein the risk source category includes personnel behavior risk, equipment state risk, environmental abnormal risk and regional location risk; Calculating the risk occurrence probability of each risk source category based on the feature vector set corresponding to each risk source category and the historical accident features in the risk database; Predicting the accident consequence severity of each risk source category based on a pre-set consequence evaluation rule; Statistically obtaining the risk exposure time of the extracted risk source category from the feature vector set; Calculating the independent risk value of each risk source using a weighted algorithm based on the risk occurrence probability, the accident consequence severity and the risk exposure time; Aggregating the independent risk values of all risk source categories and calculating the comprehensive risk evaluation value using a pre-set comprehensive evaluation function. 6.The smart glasses-based power construction safety risk point grading management method according to claim 1, characterized in that, Mapping the comprehensive risk evaluation value based on a pre-set risk level threshold interval, specifically: When the comprehensive risk evaluation value falls within the first risk level threshold interval, it is mapped to a low risk level; When the comprehensive risk evaluation value falls within the second risk level threshold interval, it is mapped to a general risk level; When the comprehensive risk evaluation value falls into the third risk level threshold interval, it is mapped to a medium risk level; When the comprehensive risk evaluation value falls into the fourth risk level threshold interval, it is mapped to a high risk level. 7.The smart glasses-based power construction safety risk point grading management method according to claim 1, characterized in that, Based on different risk levels, corresponding hierarchical warning instructions are generated, specifically: When the risk level is low risk, a first-level warning instruction is generated, a soft text prompt icon is displayed on the edge of the field of view through the display unit of the intelligent glasses, and a soft voice reminder is played once through the audio unit; When the risk level is general risk, a second-level warning instruction is generated, a prominent text warning information is displayed in the center of the field of view through the display unit, a first-level voice warning is played in a loop through the audio unit, and a first-level vibration is generated through the tactile feedback unit; When the risk level is medium risk, a third-level warning instruction is generated, detailed risk description and disposal suggestions are displayed in a flashing manner with a highlight color through the display unit, a second-level voice warning is played through the audio unit, and a second-level vibration is generated through the tactile feedback unit; When the risk level is high risk, a fourth-level warning instruction is generated, a red emergency warning information and emergency evacuation guide are displayed in a full-screen overlay manner through the display unit, a third-level voice warning is played through the audio unit, and a third-level vibration is generated through the tactile feedback unit. 8.The smart glasses-based power construction safety risk point grading management method of claim 1, wherein, The generation process of the hierarchical warning instruction also includes self-adaptive adjustment of the display position and prompt intensity of the warning content in combination with the real-time working state and position information of the construction personnel. 9.The smart glasses-based power construction safety risk point grading management method of claim 1, wherein, It also includes recording and uploading the full-process data of the current risk identification event to a remote safety management platform for updating the consequence evaluation rules and risk database, optimizing the calculation process of the comprehensive risk evaluation value, and generating a safety management report.

10. The power construction safety risk point grading management system based on smart glasses, characterized in that, The intelligent glasses include an image acquisition unit, an environment sensor unit, a positioning unit, a display unit, an audio unit, and a tactile feedback unit, and the system includes: A data acquisition module acquires multi-modal data of the power construction site using the intelligent glasses, including construction scene images, construction personnel operation behavior videos, environment data, and position data; A feature extraction module extracts risk target features and behavior pattern features from the construction scene images and construction personnel operation behavior videos based on a pre-trained deep learning model; and extracts abnormal environment features from the environment data; A risk assessment module generates a feature vector set in combination with the risk target features, the behavior pattern features, the abnormal environment features, and the position data; and calculates a comprehensive risk evaluation value based on the feature vector set; An instruction generation module maps the comprehensive risk evaluation value based on pre-set risk level threshold intervals to obtain a corresponding risk level; and generates corresponding hierarchical warning instructions based on different risk levels. An instruction generation module maps the comprehensive risk evaluation value based on pre-set risk level threshold intervals to obtain a corresponding risk level; and generates corresponding hierarchical warning instructions based on different risk levels.

Images