A power grid worker vital sign data secure transmission method and system

Abstract

This invention provides a method and system for secure transmission of vital sign data of power grid workers, belonging to the field of power system security technology. The method includes: receiving a first data packet sent by a smart wearable device, the first data packet containing an initial transmission count field and vital sign data of power grid workers; incrementing the value of the initial transmission count field to generate an updated transmission count field, and generating a second data packet based on the updated transmission count field and the vital sign data; sending the second data packet to a cloud platform; the cloud platform incrementing the transmission count field of the second data packet again and performing a transmission count verification based on this field; if the verification result does not meet a preset value, the second data packet is determined to be abnormal data and discarded. By implementing this invention, the problem of inaccurate data transmission of private information such as vital sign data of power grid workers due to security risks such as eavesdropping, tampering, or unauthorized access during transmission can be solved.

Description

Technical Field

[0001] This invention relates to the field of power system safety technology, specifically to a method and system for securely transmitting vital sign data of power grid workers. Background Technology

[0002] With the development of smart grids and digital operation and maintenance, power grid operations are gradually evolving towards informatization and intelligence. Especially in complex scenarios such as high-voltage transmission, substation operation and maintenance, and offshore wind power, the safety of workers is becoming increasingly important. Power grid operating environments are typically characterized by high voltage, high risk, and dispersed operations. If sudden health abnormalities occur to on-site personnel (such as abnormal heart rate, excessive fatigue, or sudden illness), and these are not detected and responded to in a timely manner, safety accidents can easily occur. Therefore, real-time collection of vital signs data such as heart rate, blood oxygen, and body temperature from power grid workers using wearable devices, and transmitting the relevant data to a back-end monitoring platform, has become an important means of improving the safety of power grid operations.

[0003] However, vital signs and identity information data have strong privacy attributes and are easily subject to security risks such as eavesdropping, tampering or unauthorized access during transmission. Without an effective authentication mechanism, it is impossible to determine whether the received data is accurate, which can easily lead to data errors and omissions. Summary of the Invention

[0004] A method and system for secure transmission of vital signs data of power grid workers is proposed. By introducing an authentication mechanism based on the transmission count technical field, the system can solve the problem of data inaccuracy caused by security risks such as eavesdropping, tampering, or unauthorized access to the vital signs data of power grid workers during transmission.

[0005] A method for securely transmitting vital sign data of power grid workers, applicable to user mobile terminals, includes: Receive the first data packet transmitted by the smart wearable device: wherein the first data packet contains an initial transmission count field and the physical condition data of the power grid worker; The value of the initial transmission count field in the first data packet is incremented to generate an updated transmission count field. Based on the updated transmission count field and the physical condition data of the power grid workers, a second data packet is generated. The second data packet is sent to the cloud platform. After receiving the second data packet, the cloud platform increments the value of the field in the first updated transmission count field to generate a second updated transmission count field. The transmission count of the second updated transmission count field is then checked. If the transmission count does not meet the preset value, the second data packet is treated as abnormal data and discarded.

[0006] Furthermore, before incrementing the value of the initial transmission count field in the first data packet, the following is also included: The system acquires the worker's identity information, facial video recording, and location information; performs facial anti-spoofing detection on the identity information based on the facial video recording; if the facial anti-spoofing detection passes, the value of the initial transmission count field in the first data packet is incremented; if the facial anti-spoofing detection fails, the system reacquires the facial video recording and performs facial anti-spoofing detection.

[0007] Furthermore, based on the updated transmission count field and the physical condition data of the power grid workers, a second data packet is generated, including: The facial video recordings, identity information, location information, and vital signs data of power grid workers are linked to generate associated data; The associated data is encapsulated with the updated transmission count technology field to generate a second data packet.

[0008] Furthermore, facial recognition technology is used to verify identity information based on facial video recordings, including: The face video is input into the face anti-spoofing detection model so that the face anti-spoofing detection model can determine whether the face video is a static image or a video playback. If the face recording is determined to be a static image or video playback, the face anti-spoofing detection result is "fail"; if the face recording is determined not to be a static image or video playback, the face anti-spoofing detection result is "pass".

[0009] Furthermore, if the number of transmissions meets the preset value, vital sign data, facial video recordings, identity information, and location information are extracted from the second data packet; The face video is input into the face analysis model so that the face analysis model can output emotion labels and physiological index prediction results based on the face video; Based on location information, obtain the operating environment parameters and the type of operation; The model inputs emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, work environment parameters, and job type into the safety assessment model so that the safety assessment model can generate risk identification results for power grid workers during their work.

[0010] Furthermore, the face analysis model includes an emotion recognition sub-model and a physiological indicator prediction sub-model; The facial analysis model outputs emotion labels and physiological indicator predictions based on facial video recordings, including: Select several key frames from the face video recording; Based on the emotion recognition sub-model, emotion feature values ​​are extracted from each face keyframe, and the extracted emotion feature values ​​are classified to generate emotion recognition labels. Physiological indicators are predicted for each face keyframe based on the physiological indicator prediction sub-model, and the physiological indicator prediction results are obtained.

[0011] Furthermore, emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, and location information are input into the safety assessment model; enabling the safety assessment model to generate risk identification results for power grid workers during operations based on facial video recordings, including: Emotion recognition labels, physiological indicator prediction results, and vital sign data are classified as the first type of input data; based on identity information, corresponding historical health data records are extracted from the historical database and classified as the second type of input data; based on location information, work environment parameters and work type are obtained and classified as the third type of input data; after preprocessing the first, second, and third types of input data, the input feature vector is obtained. The input feature vector is fed into the safety assessment model, which, after fusion calculation, outputs the risk probability of power grid workers during their operations. If the risk probability is greater than a preset risk probability threshold, it is determined that there is a safety risk when the power grid worker is working; otherwise, it is determined that there is no safety risk when the power grid worker is working.

[0012] Based on the above method embodiments, the present invention provides corresponding system embodiments.

[0013] A secure transmission system for vital signs data of power grid workers includes: a smart wearable device, a user mobile terminal, and a cloud platform; The smart wearable device is used to transmit a first data packet to a user's mobile terminal; wherein, the first data packet contains an initial transmission count field and the physical condition data of the power grid worker; The user mobile terminal is used to receive the first data packet transmitted by the smart wearable device, increment the value of the initial transmission count field in the first data packet to generate an updated transmission count field, generate a second data packet based on the updated transmission count field and the physical condition data of the power grid worker, and transmit the second data packet to the cloud platform. The cloud platform is used to increment the value of the transmission count field after receiving the second data packet, generate a second updated transmission count field, and perform transmission count verification on the second updated transmission count field. If the transmission count does not meet the preset value, the second data packet is treated as abnormal data and discarded.

[0014] Furthermore, the user mobile terminal is also used to acquire the worker's identity information, facial video recording, and location information; to perform face anti-spoofing detection on the identity information based on the facial video recording; if the face anti-spoofing detection passes, to increment the value of the initial transmission count field in the first data packet; if the face anti-spoofing detection fails, to reacquire the facial video recording and perform face anti-spoofing detection.

[0015] Furthermore, the user's mobile terminal generates a second data packet based on the updated transmission count field and the physical condition data of the power grid workers, including: The facial video recordings, identity information, location information, and vital signs data of power grid workers are linked to generate associated data; The associated data is encapsulated with the updated transmission count technology field to generate a second data packet.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method for securely transmitting vital sign data of power grid workers. The method introduces a transmission count field into the data packets transmitted between a smart wearable device, a user's mobile terminal, and a cloud platform. As the data packet passes through each device, the transmission count corresponding to this field increases. Upon transmission to the cloud platform, the transmission count field is verified against a preset value to ensure the security of vital sign data transmission. If they do not match, it indicates that the data has not passed through the smart wearable device, user's mobile terminal, and cloud platform sequentially, and therefore there is a possibility of interception and tampering. In this case, the data is directly discarded as abnormal data to ensure the accuracy of the privacy data. Attached Figure Description

[0017] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a method for securely transmitting vital signs data of power grid workers according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of a secure transmission system for the vital signs data of power grid workers provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] To address the issue of data inaccuracy caused by security risks such as eavesdropping, tampering, or unauthorized access during the transmission of vital signs data and other private information of power grid workers, embodiments of the present invention provide a method and system for secure transmission of vital signs data of power grid workers. This method and system can solve the problem that, in the prior art, the lack of effective encryption and authentication mechanisms during the transmission of vital signs and identity information data of power grid workers exposes the data to information security risks such as eavesdropping, tampering, or unauthorized access.

[0021] An embodiment of the present invention provides a method for secure transmission of vital sign data of power grid workers, comprising at least the following steps: Step S1: Receive the first data packet transmitted by the smart wearable device: wherein the first data packet contains an initial transmission count field and the physical condition data of the power grid worker; Specifically, vital signs data are collected from power grid workers using smart wearable devices. These vital signs data include information such as heart rate, body temperature, blood oxygen saturation, and exercise status. This vital signs data is packaged and encapsulated, and a transmission count field is introduced into the packaged data packet and initialized. The initial value of the transmission count field is 1, that is, the initial transmission count field is 1. The data packet with the transmission count field added is then sent to the user's mobile terminal.

[0022] In this invention, the collection of vital signs data of power grid workers through smart wearable devices enables the acquisition of vital signs data of power grid workers without affecting their normal work, thus achieving non-invasive inspection. At the same time, in high-risk outdoor work environments such as power grid inspections, the collection of vital signs data of power grid workers through smart wearable devices enables remote monitoring and real-time feedback on the workers' physical condition, achieving a better level of health management for power grid workers.

[0023] Step S2: Increment the value of the initial transmission count field in the first data packet to generate an updated transmission count field. Based on the updated transmission count field and the physical condition data of the power grid workers, generate the second data packet. Specifically, after receiving the first data packet, the first data packet is decapsulated to obtain the physical condition data of the power grid workers and the initial transmission count field within the first data packet; Next, the initial transmission count field is incremented, for example, by simply adding 1; thus, the transmission count field after the first update is obtained. Then, based on the successively updated transmission count field and the physical condition data of the power grid workers, a second data packet is generated.

[0024] In a preferred embodiment, before incrementing the value of the initial transmission count field in the first data packet, the method further includes: acquiring the worker's identity information, facial video recording, and location information; performing face verification based on the facial video recording; if the face verification passes, incrementing the value of the initial transmission count field in the first data packet; if the face verification fails, re-acquiring the facial video recording and performing face verification again. Specifically, in this embodiment, before incrementing the value of the initial transmission count field, the user's mobile terminal internal program first obtains the power grid worker's identity information, facial video recording, and location information. Based on the facial video recording, a face verification test is performed. If the test passes, it indicates that the current facial video recording is being collected in real-time and there is no identity fraud issue; in this case, the value of the initial transmission count field in the first data packet is incremented. If the test fails, it indicates that the current facial video recording is not being collected in real-time and there is a risk of identity fraud; in this case, the facial video recording needs to be collected again until the face verification test passes. If the face verification test fails after a preset number of attempts (e.g., five), data transmission is stopped.

[0025] In this embodiment, verifying the collected facial video recordings before updating the value of the transmission count field can effectively prevent the identity of power grid workers from being impersonated and ensure that the collected identity and location information has high credibility. At the same time, verifying the facial video recordings can effectively prevent the server from being attacked by automated attacks and malicious data injection, thereby improving the security of the cloud server.

[0026] In a preferred embodiment, facial identity verification based on facial video recording includes: The face video is input into the face anti-spoofing detection model so that the face anti-spoofing detection model can determine whether the face video is a static image or a video playback. If the face recording is determined to be a static image or video playback, the face anti-spoofing detection result is "fail"; if the face recording is determined not to be a static image or video playback, the face anti-spoofing detection result is "pass".

[0027] In this embodiment, a face anti-spoofing detection model is used to verify whether the face video recording is a static image or video playback to determine if the face video recording is not real-time recording. If the face anti-spoofing detection model determines that the face video recording is a static image or video playback, it is determined that the face video recording is real-time recording, and the face anti-spoofing detection result is "failed." Otherwise, the face anti-spoofing detection result is "passed." Detecting whether the current face video recording is a static image or video playback based on a face anti-spoofing detection model is existing technology and will not be elaborated upon here. In a preferred embodiment, a second data packet is generated based on an updated transmission count field and the physical condition data of the power grid workers, including: The facial video recordings, identity information, location information, and vital signs data of power grid workers are linked to generate associated data; The associated data is encapsulated with the updated transmission count technology field to generate a second data packet.

[0028] Specifically, after the smart wearable device generates vital sign data, it configures a corresponding identity identifier based on the user's identity. If the user's mobile terminal passes the face anti-counterfeiting detection, the identity identifier is matched with the user's identity information. After successful matching, the identity information, facial video recording, location information, and vital sign data are associated to obtain the aforementioned associated data. The associated data and the updated transmission count field (e.g., 2) are packaged and encapsulated to generate the aforementioned second data packet, which is then sent to the cloud platform.

[0029] Step S3: Send the second data packet to the cloud platform so that after receiving the second data packet, the cloud platform increments the value of the transmission count field after the first update to generate the transmission count field after the second update. The transmission count field after the second update is checked. If the transmission count does not meet the preset value, the second data packet is treated as abnormal data and discarded. Specifically, after receiving the data packet sent by the mobile terminal, the cloud platform increments the value of the transmission count field in the data packet, for example, by directly adding 1, to obtain the transmission count field after the second update. Then, the value of the transmission count field after the second update is verified; If the value of the transmission count field after the second update is not equal to the preset value (the illustrative preset value is 3), the data packet is considered abnormal and discarded; otherwise, the data packet is considered normal. When the data packet is normal, the data packet is unblocked.

[0030] In this invention, by verifying the value of the transmission count field, the following effects can be achieved: 1. By determining the number of devices the data packet passes through by the transmission count field, illegal forwarding or illegal tampering of the data packet during transmission is avoided, ensuring the legitimacy of the data source; 2. After verifying the legitimacy of the data packet, unpacking can reduce invalid calculations and improve the data processing efficiency of the cloud platform.

[0031] In a preferred embodiment, if the number of transmissions meets a preset value, vital sign data, facial video recording, identity information, and location information are extracted from the second data packet; the facial video recording is input into the face analysis model so that the face analysis model outputs emotion tags and physiological indicator prediction results based on the facial video recording; Based on the location information, the working environment parameters and the type of work are obtained; The model inputs emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, work environment parameters, and job type into the safety assessment model so that the safety assessment model can generate risk identification results for power grid workers during their work.

[0032] Specifically, vital sign data, facial video recordings, identity information, and location information are extracted from the verified data packets; The face video is input into the face analysis model; the face analysis model extracts feature values ​​from the face video and adjusts the internal parameters of the model, and finally outputs the emotion label and physiological index prediction results corresponding to the face video; Based on location information, the cloud platform queries the database to obtain the operating environment parameters and the type of operation; The safety assessment model is input into the emotion recognition tags, physiological indicator prediction results, vital sign data, work environment parameters, and job type. The safety assessment model extracts features from the emotion recognition tags, physiological indicator prediction results, vital sign data, work environment parameters, and job type, and adjusts the internal parameters of the model to finally output the risk identification results of power grid workers during their work.

[0033] In a preferred embodiment, the face analysis model includes an emotion recognition sub-model and a physiological indicator prediction sub-model. The face analysis model outputs emotion labels and physiological indicator prediction results based on the face video recording, including: selecting several key frames from the face video recording; extracting emotion feature values ​​from each face key frame based on the emotion recognition sub-model; classifying the extracted emotion feature values; and generating emotion recognition labels. Physiological indicators are predicted for each face keyframe based on the physiological indicator prediction sub-model, and the physiological indicator prediction results are obtained.

[0034] Specifically, the face analysis model consists of a physiological indicator prediction sub-model and an emotion recognition sub-model; A short video of a frontal face, ranging from 15 to 30 seconds, is used as the model input. Preprocessing is performed first: keyframes are extracted at a rate of 10 frames per second. The MTCNN face detection algorithm is used to locate and align the face region in each frame. Each keyframe is then input into the emotion recognition sub-model and the physiological indicator prediction sub-model, respectively. Regarding the application stage of the face analysis model, the physiological index prediction sub-model takes each key frame as input and selects the cheek area of ​​each key frame as the region of interest. It uses the independent component analysis algorithm to separate the photoplethysmography signal from the RGB color channel of the image sequence, and then performs time-series filtering and signal amplification processing to calculate physiological indicators such as respiratory rate. Blood pressure estimation was achieved using a regression sub-model based on pulse wave transit time, while body temperature estimation was accomplished by analyzing chromaticity changes and infrared features in a specific ROI region. The emotion recognition sub-model takes each keyframe as input, uses a pre-trained ResNet-50 network to extract facial features for each frame, and then classifies the temporal features through a long short-term memory network to output emotion labels such as calm, excited, sad, or joy. The calculation of physiological indicators based on photoplethysmography (PPG) signals and the output based on the ResNet-50 network and long short-term memory network are both existing technologies and will not be elaborated upon here. Regarding the training phase of the face analysis model, both the physiological indicator prediction sub-model and the emotion recognition sub-model adopt supervised deep learning training methods. The physiological indicator prediction sub-model uses synchronously collected face videos and real values ​​measured by medical equipment as the training set, and optimizes the model parameters by minimizing the mean squared error loss function. The emotion recognition sub-model is trained using the AffectNet public dataset and adopts the cross-entropy loss function.

[0035] In this embodiment, by introducing a face analysis model, it is possible to predict vital signs (such as respiratory rate) collected when face video recordings cannot be directly accessed, providing richer information for subsequent safety assessment models and expanding the dimensions of available information; it is also possible to obtain more vital sign information of power grid workers without increasing hardware equipment, thereby reducing the maintenance costs of hardware equipment.

[0036] In a preferred embodiment, emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, work environment parameters, and job type are input into a safety assessment model. This enables the safety assessment model to generate risk identification results for power grid workers based on facial video recordings. The process includes: classifying emotion recognition tags, physiological indicator prediction results, and vital sign data as a first type of input data; extracting corresponding historical health data records from a historical database based on identity information and classifying these historical health data records as a second type of input data; obtaining work environment parameters and job type based on location information and classifying them as a third type of input data; and preprocessing the first, second, and third types of input data to obtain an input feature vector. The input feature vector is fed into the safety assessment model. After fusion calculation, the safety assessment model outputs the risk probability of power grid workers during their work. If the risk probability is greater than the preset risk probability threshold, it is determined that there is a safety risk when power grid workers are working; otherwise, it is determined that there is no safety risk when power grid workers are working.

[0037] Specifically, in the application phase of the safety assessment model, identity information, vital sign data, emotion tags, predicted vital sign values, work environment parameters, and job type are used as inputs to the preprocessing module of the safety assessment model. The input data is then categorized as follows: The first category consists of real-time physiological indicators and emotion tags output by the face analysis model, such as heart rate, respiratory rate, estimated body temperature, and emotional state; the second category consists of historical physical examination data and health record information retrieved from the database, including past physical examination indicators, chronic disease history, and abnormal health records; and the third category consists of work environment parameters and job type information obtained based on location information, such as ambient temperature, work area identification, and job type code. The three types of data mentioned above undergo data processing before being input into the model, including outlier removal, missing value completion, and standardization. Continuous variables are scaled using normalization or standardization methods, while categorical variables are converted into numerical form using encoding methods. After processing, the various features are concatenated in a uniform order to form a fixed-dimensional feature vector, which serves as the input to the health and safety risk assessment model. After feature preprocessing (including missing value imputation, normalization, and feature encoding) and vectorization, a unified-dimensional real-time feature vector is obtained and input into the trained XGBoost security assessment model. The model consists of multiple decision trees built using a gradient boosting strategy. Input samples first enter the first decision tree, and are judged layer by layer according to the feature partitioning conditions of each node (such as feature value range or threshold comparison). Following the corresponding branch path, the sample reaches a leaf node, thus obtaining the tree's output value for the current sample (i.e., the local risk score).

[0038] Subsequently, the samples are input into each subsequent decision tree in the same manner. Each tree learns from the residuals of the previous model, characterizing and correcting the risk of the sample from different perspectives. The outputs of all decision trees are fused using an additive model, that is, the output values ​​of each tree are weighted and accumulated (the weights are adaptively learned during model training and can be scaled by the learning rate) to obtain a comprehensive score for the sample. For binary or multi-class classification tasks, this comprehensive score is usually further mapped using a sigmoid function or a softmax function to transform it into the corresponding risk prediction probability value.

[0039] After obtaining the final predicted probability, it is compared with a pre-set risk threshold: when the predicted probability is higher than the threshold, it is determined to be a high-risk state; when the predicted probability is lower than the threshold, it is determined to be a low-risk state. A green label indicates that the operation is permitted, while a red label indicates that the operation is prohibited. The corresponding prediction confidence level is also provided to assist the site management system in conducting operation access control. The prediction probability is output based on the XGBoost model, which is existing technology and will not be elaborated upon here. During the training phase of the safety assessment model, a training dataset is first constructed. The training dataset consists of samples and labels. The samples include identity information, vital sign data, emotion labels, predicted vital sign values, work environment parameters, and work types. The labels are manually assigned to the samples based on the health status and safety records during the work process, forming two categories of labels: "green (work permitted)" and "red (work prohibited)". During training, the model uses the cross-entropy loss function as the optimization objective and generates multiple decision trees in each round through the gradient boosting strategy. In each round of training, the sample weights are updated based on the prediction error of the previous round of the model, so that the subsequent generated decision trees pay more attention to samples that are difficult to classify correctly. Each decision tree selects the optimal feature and split threshold to divide the samples, thereby gradually learning the mapping relationship between input features and health risk levels. The outputs of multiple decision trees are weighted and summed to form the final prediction function.

[0040] In this embodiment, by introducing a safety assessment model, multi-source data such as worker vital signs, behavioral characteristics, and environmental information are fused and modeled, which can fully explore the nonlinear relationships between various features, thereby achieving an accurate assessment of the safe working status. By leveraging the robustness and computational efficiency of the security assessment model, the impact of noisy data on the assessment results can be effectively reduced, and real-time risk analysis and early warning can be supported, thereby improving the system's intelligence level and security capabilities.

[0041] The above-described method embodiments, by introducing a transmission count technical field into the data packet, can effectively reduce the risk of illegal interception or tampering of the aforementioned private information during transmission while collecting the physical characteristics and identity information of power grid workers, thus effectively preventing the leakage of power grid workers' privacy.

[0042] Based on the above method embodiments, the present invention provides corresponding system embodiments.

[0043] An embodiment of the present invention provides a secure transmission system for the vital signs data of power grid workers, comprising: The smart wearable device is used to transmit a first data packet to a user's mobile terminal; wherein, the first data packet contains an initial transmission count field and the physical condition data of the power grid worker; The user mobile terminal is used to receive the first data packet transmitted by the smart wearable device, increment the value of the initial transmission count field in the first data packet to generate an updated transmission count field, generate a second data packet based on the updated transmission count field and the physical condition data of the power grid worker, and transmit the second data packet to the cloud platform. The cloud platform is used to increment the value of the transmission count field after receiving the second data packet, generate a second updated transmission count field, and perform transmission count verification on the second updated transmission count field. If the transmission count does not meet the preset value, the second data packet is treated as abnormal data and discarded.

[0044] Preferably, the user mobile terminal is further used to acquire the worker's identity information, facial video recording, and location information; perform face anti-counterfeiting detection on the identity information based on the facial video recording; if the face anti-counterfeiting detection passes, increment the value of the initial transmission count field in the first data packet; if the face anti-counterfeiting detection fails, reacquire the facial video recording and perform face anti-counterfeiting detection. The user's mobile terminal generates a second data packet based on the updated transmission count field and the physical characteristics data of the power grid worker. This includes: associating the power grid worker's facial video recording, identity information, location information, and physical characteristics data to generate associated data; and encapsulating the associated data with the updated transmission count field to generate the second data packet.

[0045] In a preferred embodiment, the user's mobile terminal performs facial identity verification based on facial video recording, including: The face video is input into the face anti-spoofing detection model so that the face anti-spoofing detection model can determine whether the face video is a static image or a video playback. If the face recording is determined to be a static image or video playback, the face anti-spoofing detection result is "fail"; if the face recording is determined not to be a static image or video playback, the face anti-spoofing detection result is "pass".

[0046] In a preferred embodiment, the cloud platform, provided that the number of transmissions meets a preset value, is further configured to extract vital sign data, facial video recordings, identity information, and location information from the second data packet; and input the facial video recordings into a face analysis model so that the face analysis model outputs emotion tags and physiological indicator prediction results based on the facial video recordings. Based on the location information, the working environment parameters and the type of work are obtained; The model inputs emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, work environment parameters, and job type into the safety assessment model so that the safety assessment model can generate risk identification results for power grid workers during their work.

[0047] In a preferred embodiment, the face analysis model includes an emotion recognition sub-model and a physiological indicator prediction sub-model. The face analysis model outputs emotion labels and physiological indicator prediction results based on the face video recording, including: selecting several key frames from the face video recording; extracting emotion feature values ​​from each face key frame based on the emotion recognition sub-model; classifying the extracted emotion feature values; and generating emotion recognition labels. Physiological indicators are predicted for each face keyframe based on the physiological indicator prediction sub-model, and the physiological indicator prediction results are obtained.

[0048] In a preferred embodiment, the cloud platform inputs emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, work environment parameters, and job type into a safety assessment model; so that the safety assessment model generates risk identification results for power grid workers based on facial video recordings, including: classifying emotion recognition tags, physiological indicator prediction results, and vital sign data as a first type of input data; extracting corresponding historical health data records from a historical database based on identity information and classifying these historical health data records as a second type of input data; obtaining work environment parameters and job type based on location information and classifying them as a third type of input data; and preprocessing the first, second, and third types of input data to obtain an input feature vector. The input feature vector is fed into the safety assessment model. After fusion calculation, the safety assessment model outputs the risk probability of power grid workers during their work. If the risk probability is greater than the preset risk probability threshold, it is determined that there is a safety risk when power grid workers are working; otherwise, it is determined that there is no safety risk when power grid workers are working.

[0049] Specifically, the smart wearable device collects vital signs data from power grid workers, including information such as heart rate, body temperature, blood oxygen saturation, and exercise status. This vital signs data is packaged and encapsulated, and a transmission count field is introduced into the packaged data packet and initialized. The initial value of the transmission count field is 1, that is, the initial transmission count field is 1. The data packet with the transmission count field added is sent to the user's mobile terminal.

[0050] In this invention, the collection of vital signs data of power grid workers through smart wearable devices enables the acquisition of vital signs data of power grid workers without affecting their normal work, thus achieving non-invasive inspection. At the same time, in high-risk outdoor work environments such as power grid inspections, the collection of vital signs data of power grid workers through smart wearable devices enables remote monitoring and real-time feedback on the workers' physical condition, achieving a better level of health management for power grid workers.

[0051] After receiving the first data packet, the user's mobile terminal decrypts the first data packet and obtains the physical condition data of the power grid workers and the initial transmission count field from the first data packet. The user's mobile terminal then increments the initial transmission count field, for example, by simply adding 1; thus obtaining the updated transmission count field. Then, based on the successively updated transmission count field and the physical condition data of the power grid workers, a second data packet is generated; Preferably, before the user's mobile terminal increments the value of the initial transmission count field, the process includes: first obtaining the power grid worker's identity information, facial video recording, and location information through the user's mobile terminal's internal program; Face anti-spoofing detection is performed based on face video recording. If the detection passes, it means that the current face video is collected in real time and there is no identity fraud problem. At this time, the value of the initial transmission count field in the first data packet is incremented. If the verification fails, it indicates that the current facial video recording is not being captured in real time, posing a risk of identity theft. In this case, the facial video recording needs to be captured again until the face verification detection passes. If the face verification detection fails after a preset number of attempts (e.g., five), data transmission will be stopped.

[0052] Verifying the collected facial video footage before updating the transmission count field effectively prevents the identity of power grid workers from being impersonated, ensuring the high credibility of the collected identity and location information. At the same time, verifying the facial video footage can effectively prevent the server from being attacked by automated systems and malicious data injection, improving the security of the cloud server.

[0053] When the user's mobile terminal passes the face anti-counterfeiting detection, the identity identifier is matched with the user's identity information. After successful matching, the identity information, facial video, location information, and vital sign data are associated to obtain the aforementioned associated data. The associated data and the updated transmission count field (e.g., 2) are packaged and encapsulated to generate the aforementioned second data packet, and the second data packet is sent to the cloud platform.

[0054] After receiving the data packet sent by the mobile terminal, the cloud platform increments the value of the transmission count field in the data packet. For example, it directly adds 1 to obtain the transmission count field after the second update. The cloud platform verifies the value of the transmission count field after the second update. If the value of the transmission count field after the second update is not equal to the preset value (the illustrative preset value is 3), the data packet is considered abnormal and discarded; otherwise, the data packet is considered normal. When the data packet is normal, the data packet is unblocked.

[0055] In this invention, by verifying the value of the transmission count field, the following effects can be achieved: 1. By determining the number of devices the data packet passes through by the transmission count field, illegal forwarding or illegal tampering of the data packet during transmission is avoided, ensuring the legitimacy of the data source; 2. After verifying the legitimacy of the data packet, unpacking can reduce invalid calculations and improve the data processing efficiency of the cloud platform.

[0056] The cloud platform extracts vital signs data, facial video recordings, identity information, and location information from the verified data packets; The face video is input into the face analysis model; the face analysis model extracts feature values ​​from the face video and adjusts the internal parameters of the model, and finally outputs the emotion label and physiological index prediction results corresponding to the face video; Based on location information, the cloud platform queries the database to obtain the operating environment parameters and the type of operation; The safety assessment model is input into the emotion recognition tags, physiological indicator prediction results, vital sign data, work environment parameters, and job type. The safety assessment model extracts features from the emotion recognition tags, physiological indicator prediction results, vital sign data, work environment parameters, and job type, and adjusts the internal parameters of the model to finally output the risk identification results of power grid workers during their work.

[0057] Preferably, the face analysis model consists of a physiological indicator prediction sub-model and an emotion recognition sub-model; A short video of a frontal face, ranging from 15 to 30 seconds, is used as the model input. Preprocessing is performed first: keyframes are extracted at a rate of 10 frames per second. The MTCNN face detection algorithm is used to locate and align the face region in each frame. Each keyframe is then input into the emotion recognition sub-model and the physiological indicator prediction sub-model, respectively. Regarding the application stage of the face analysis model, the physiological index prediction sub-model takes each key frame as input and selects the cheek area of ​​each key frame as the region of interest. It uses the independent component analysis algorithm to separate the photoplethysmography signal from the RGB color channel of the image sequence, and then performs time-series filtering and signal amplification processing to calculate physiological indicators such as respiratory rate. Blood pressure estimation was achieved using a regression sub-model based on pulse wave transit time, while body temperature estimation was accomplished by analyzing chromaticity changes and infrared features in a specific ROI region. The emotion recognition sub-model takes each keyframe as input, uses a pre-trained ResNet-50 network to extract facial features for each frame, and then classifies the temporal features through a long short-term memory network to output emotion labels such as calm, excited, sad, or joy. The calculation of physiological indicators based on photoplethysmography (PPG) signals and the output based on the ResNet-50 network and long short-term memory network are both existing technologies and will not be elaborated upon here. Regarding the training phase of the face analysis model, both the physiological index prediction sub-model and the emotion recognition sub-model adopt supervised deep learning training methods; the physiological index prediction sub-model uses synchronously collected face videos and real values ​​measured by medical equipment as the training set, and optimizes the model parameters by minimizing the mean square error loss function. The emotion recognition sub-model was trained using the AffectNet public dataset and employed the cross-entropy loss function.

[0058] This invention introduces a face analysis model, which can predict vital signs (such as respiratory rate) when face video recordings cannot be directly accessed, providing richer information for subsequent safety assessment models and expanding the dimensions of available information; it can also acquire more vital signs information of power grid workers without increasing hardware equipment, thus reducing the maintenance costs of hardware equipment.

[0059] In the application phase of the safety assessment model, identity information, vital sign data, emotion tags, predicted vital sign values, work environment parameters, and job type are used as inputs to the preprocessing module of the safety assessment model. The input data is categorized as follows: The first category is real-time physiological indicators and emotion tags output by the face analysis model, such as heart rate, respiratory rate, estimated body temperature, and emotional state; the second category is historical physical examination data and health record information retrieved from the database, including past physical examination indicators, chronic disease history, and abnormal health records; the third category is work environment parameters and job type information obtained based on location information, such as ambient temperature, work area identification, and job type code. Before inputting the above three types of data into the model, data processing is performed, including outlier removal, missing value completion, and standardization. Continuous variables are normalized or standardized for scaling, while categorical variables are converted to numerical form using encoding methods. After processing, the various features are concatenated in a unified order to form a fixed-dimensional feature vector, which serves as the input to the health and safety risk assessment model. After feature preprocessing (including missing value imputation, normalization, and feature encoding) and vectorization, a unified-dimensional real-time feature vector is obtained and input into the trained XGBoost security assessment model. The model consists of multiple decision trees built using a gradient boosting strategy. Input samples first enter the first decision tree, and are judged layer by layer according to the feature partitioning conditions of each node (such as feature value range or threshold comparison). The sample then follows the corresponding branch path to reach a leaf node, thus obtaining the tree's output value for the current sample (i.e., the local risk score).

[0060] Subsequently, the samples are input into each subsequent decision tree in the same manner. Each tree learns from the residuals of the previous model, characterizing and correcting the risk of the sample from different perspectives. The outputs of all decision trees are fused using an additive model, that is, the output values ​​of each tree are weighted and accumulated (the weights are adaptively learned during model training and can be scaled by the learning rate) to obtain a comprehensive score for the sample. For binary or multi-class classification tasks, this comprehensive score is usually further mapped using a sigmoid function or a softmax function to transform it into the corresponding risk prediction probability value.

[0061] After obtaining the final predicted probability, it is compared with a pre-set risk threshold: when the predicted probability is higher than the threshold, it is determined to be a high-risk state; when the predicted probability is lower than the threshold, it is determined to be a low-risk state. A green label indicates that the operation is permitted, while a red label indicates that the operation is prohibited. The corresponding prediction confidence level is also provided to assist the site management system in conducting operation access control. The prediction probability is output based on the XGBoost model, which is existing technology and will not be elaborated upon here. During the training phase of the safety assessment model, a training dataset is first constructed. The training dataset consists of samples and labels. The samples include identity information, vital sign data, emotion labels, predicted vital sign values, work environment parameters, and work types. The labels are manually applied to the samples based on the health status and safety records during the operation, resulting in two types of labels: "green (operation permitted)" and "red (operation prohibited)".

[0062] During training, the model uses the cross-entropy loss function as the optimization objective and generates multiple decision trees in each round through the gradient boosting strategy. In each round of training, the sample weights are updated based on the prediction error of the previous round of the model, so that the subsequent generated decision trees pay more attention to samples that are difficult to classify correctly. Each decision tree selects the optimal feature and split threshold to divide the samples, thereby gradually learning the mapping relationship between input features and health risk levels. The outputs of multiple decision trees are weighted and summed to form the final prediction function.

[0063] By introducing a safety assessment model, multi-source data such as workers' physical characteristics, behavioral features, and environmental information are integrated and modeled, which can fully explore the nonlinear relationships between various features, thereby achieving an accurate assessment of the safe working status. By leveraging the robustness and computational efficiency of the security assessment model, the impact of noisy data on the assessment results can be effectively reduced, and real-time risk analysis and early warning can be supported, thereby improving the system's intelligence level and security capabilities.

[0064] The aforementioned system implementation, by introducing a transmission count field into the data packet, can effectively reduce the risk of illegal interception or tampering of the aforementioned private information during transmission while collecting the physical characteristics and identity information of power grid workers, thus effectively preventing the leakage of power grid workers' privacy.

[0065] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0066] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for secure transmission of power grid worker vital data, suitable for a user mobile terminal, characterized in that, include: Receive the first data packet transmitted by the smart wearable device: wherein the first data packet contains an initial transmission count field and the physical condition data of the power grid worker; The value of the initial transmission count field in the first data packet is incremented to generate an updated transmission count field. Based on the updated transmission count field and the physical condition data of the power grid workers, a second data packet is generated. The second data packet is sent to the cloud platform. After receiving the second data packet, the cloud platform increments the value of the field in the first updated transmission count field to generate a second updated transmission count field. The transmission count of the second updated transmission count field is then checked. If the transmission count does not meet the preset value, the second data packet is treated as abnormal data and discarded.

2. The method for secure transmission of vital sign data of power grid workers as described in claim 1, characterized in that, Before incrementing the value of the initial transmission count field in the first data packet, the following is also included: The system acquires the worker's identity information, facial video recording, and location information; performs facial anti-spoofing detection on the identity information based on the facial video recording; if the facial anti-spoofing detection passes, the value of the initial transmission count field in the first data packet is incremented; if the facial anti-spoofing detection fails, the system reacquires the facial video recording and performs facial anti-spoofing detection.

3. The method for secure transmission of vital sign data of power grid workers as described in claim 2, characterized in that, Based on the updated transmission count field and the physical condition data of the power grid workers, a second data packet is generated, including: The facial video recordings, identity information, location information, and vital signs data of power grid workers are linked to generate associated data; The associated data is encapsulated with the updated transmission count technology field to generate a second data packet.

4. The method for secure transmission of vital sign data of power grid workers as described in claim 3, characterized in that, Facial recognition for identity verification based on facial video recordings includes: The face video is input into the face anti-spoofing detection model so that the face anti-spoofing detection model can determine whether the face video is a static image or a video playback. If the face recording is determined to be a static image or video playback, the face anti-spoofing detection result is "fail"; if the face recording is determined not to be a static image or video playback, the face anti-spoofing detection result is "pass".

5. The method for secure transmission of vital sign data of power grid workers as described in claim 4, characterized in that, Also includes: If the number of transmissions meets the preset value, extract vital sign data, facial video recordings, identity information, and location information from the second data packet; The face video is input into the face analysis model so that the face analysis model can output emotion labels and physiological index prediction results based on the face video; Based on location information, obtain the operating environment parameters and the type of operation; The model inputs emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, work environment parameters, and job type into the safety assessment model so that the safety assessment model can generate risk identification results for power grid workers during their work.

6. The method for secure transmission of vital sign data of power grid workers as described in claim 5, characterized in that, The facial analysis model includes an emotion recognition sub-model and a physiological indicator prediction sub-model. The facial analysis model outputs emotion labels and physiological indicator predictions based on facial video recordings, including: Select several key frames from the video of the human face; Based on the emotion recognition sub-model, emotion feature values ​​are extracted from each face keyframe, and the extracted emotion feature values ​​are classified to generate emotion recognition labels. Physiological indicators are predicted for each face keyframe based on the physiological indicator prediction sub-model, and the physiological indicator prediction results are obtained.

7. The method for secure transmission of vital sign data of power grid workers as described in claim 6, characterized in that, Emotion recognition tags, physiological indicator prediction results, vital sign data, identity information, and location information are input into the security assessment model; To enable the safety assessment model to generate risk identification results for power grid workers during operations based on facial video recordings, including: Emotion recognition labels, physiological indicator prediction results, and vital sign data are classified as the first type of input data; based on identity information, corresponding historical health data records are extracted from the historical database and classified as the second type of input data; based on location information, work environment parameters and work type are obtained and classified as the third type of input data; after preprocessing the first, second, and third types of input data, the input feature vector is obtained. The input feature vector is fed into the safety assessment model, and after fusion calculation, the safety assessment model outputs the risk probability of power grid workers during operation. If the risk probability is greater than a preset risk probability threshold, it is determined that there is a safety risk when the power grid worker is working; otherwise, it is determined that there is no safety risk when the power grid worker is working.

8. A secure transmission system for vital sign data of power grid workers, characterized in that, include: Smart wearable devices, user mobile terminals, and cloud platforms; The smart wearable device is used to transmit a first data packet to a user's mobile terminal; wherein, the first data packet contains an initial transmission count field and the physical condition data of the power grid worker; The user mobile terminal is used to receive the first data packet transmitted by the smart wearable device, increment the value of the initial transmission count field in the first data packet to generate an updated transmission count field, generate a second data packet based on the updated transmission count field and the physical condition data of the power grid worker, and transmit the second data packet to the cloud platform. The cloud platform is used to increment the value of the transmission count field after receiving the second data packet, generate a second updated transmission count field, and perform transmission count verification on the second updated transmission count field. If the transmission count does not meet the preset value, the second data packet is treated as abnormal data and discarded.

9. The secure transmission system for vital sign data of power grid workers as described in claim 8, characterized in that, The user mobile terminal is also used to acquire the worker's identity information, facial video recording, and location information; to perform face anti-spoofing detection on the identity information based on the facial video recording; if the face anti-spoofing detection passes, to increment the value of the initial transmission count field in the first data packet; if the face anti-spoofing detection fails, to reacquire the facial video recording and perform face anti-spoofing detection.

10. The secure transmission system for vital sign data of power grid workers as described in claim 9, characterized in that, The user's mobile terminal generates a second data packet based on the updated transmission count field and the physical condition data of the power grid workers, including: The facial video recordings, identity information, location information, and vital signs data of power grid workers are linked to generate associated data; The associated data is encapsulated with the updated transmission count technology field to generate a second data packet.

Images