Snow mountain login safety early warning method and system based on intelligent helmet

Abstract

The invention discloses a snow mountain login safety early warning method and system based on an intelligent helmet, and relates to the technical field of machine learning, and the method comprises the steps: collecting the risk multi-dimensional data of a plurality of mountaineering routes, quantifying the basic risk state index of each node of each mountaineering route, and constructing a virtual simulation snow mountain digital twinborn model; acquiring advancing state data and environment state data under the route where the user logs in the snow mountain in real time, and evaluating a fatigue state index under a node corresponding to the route where the user logs in the snow mountain in real time; substituting the environment state data under the route where the real-time snow mountain login user is located into the virtual simulation snow mountain digital twinborn model, and quantifying a dynamic risk index of a corresponding node under the route where the real-time snow mountain login user is located; and judging whether the fatigue state index under the corresponding node of the route of the user in real-time snow mountain login meets the dynamic risk index of the corresponding node or not. The method has the advantages that the risk of man-made misjudgment is reduced, and the emergency response speed in an extreme environment is increased.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and specifically relates to a snow mountain landing safety warning method and system based on an intelligent helmet. Background Art

[0002] The multi-modal data security warning for snow mountain landing is a security monitoring and warning system based on multiple technical means, aiming to identify and analyze potential security threats in the network or data environment in real time, issue warnings through intelligent means, and realize a comprehensive security protection body for real-time monitoring, evaluation and intervention of dynamic risks during the mountain climbing process.

[0003] Due to the data acquisition and transmission delay in the existing snow mountain landing safety warning, and the low sensitivity and intelligence level of the single-modal model warning mechanism, false alarms or missed alarms are likely to occur when identifying and analyzing potential threats, resulting in untimely and ineffective security responses. This makes the built-in system of the intelligent helmet unable to comprehensively and accurately respond to complex and changeable security threats, affecting the overall security warning effect. Summary of the Invention

[0004] To solve the above technical problems, a snow mountain landing safety warning method and system based on an intelligent helmet are provided. This technical solution solves the problems of the existing snow mountain landing safety warning, such as data acquisition and transmission delay, low sensitivity and intelligence level of the single-modal model warning mechanism, easy occurrence of false alarms or missed alarms when identifying and analyzing potential threats, resulting in untimely and ineffective security responses. This makes the built-in system of the intelligent helmet unable to comprehensively and accurately respond to complex and changeable security threats, affecting the overall security warning effect.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A snow mountain landing safety warning method based on an intelligent helmet, comprising:

[0007] Determine several climbing routes for the snow mountain landing task, collect multi-dimensional risk data in several climbing routes, quantify the basic risk status indicators of each node in each climbing route, and construct a virtual simulation snow mountain digital twin model;

[0008] Based on the sensors embedded in the intelligent helmet, obtain the movement state data and environmental state data of the real-time snow mountain landing user under the current route;

[0009] According to the change trend of the movement state data and environmental state data of the real-time snow mountain landing user under the current route, evaluate the fatigue state indicators of the corresponding node of the user under the current route in the real-time snow mountain landing;

[0010] Substitute the environmental status data under the route where the real-time snow mountain landing user is located into the virtual simulation snow mountain digital twin model, perform real-time compensation for the basic risk status indicators of each node on each climbing route, and quantify the dynamic risk index of the corresponding node under the route where the real-time snow mountain landing user is located;

[0011] Judge whether the fatigue status indicator under the corresponding node of the route where the user is located during the real-time snow mountain landing meets the dynamic risk index of the corresponding node under the route where the real-time snow mountain landing user is located. If so, determine to continue logging in. If not, determine to suspend logging in.

[0012] Furthermore, based on the snow mountain to be landed, determine several climbing routes for the snow mountain landing task, and obtain the risk multi-dimensional data in several climbing routes; the multi-dimensional data of the climbing route includes: terrain data, meteorological data, and hidden danger distribution data;

[0013] Taking the supply camp in the climbing route as the division node, divide the risk multi-dimensional data in several climbing routes to obtain the risk multi-dimensional data of each node in each climbing route;

[0014] Based on the risk multi-dimensional data of each node in each climbing route, perform Z-score standardization preprocessing for numerical data and One-Hot encoding preprocessing for categorical data;

[0015] According to the Transformer architecture, train the modal classifier of the corresponding dimension type according to the risk multi-dimensional data of each node in each climbing route, and generate the risk type logical value of each node in each climbing route;

[0016] Using the Softmax function, taking the risk type logical value of each node in each climbing route as the input and taking the risk type probability distribution value of each node in each climbing route as the output.

[0017] Furthermore, according to the entropy weight method, calculate the contribution of the risk type probability distribution value of each node in each climbing route, and assign weights to the risk type probability distribution value of each node in each climbing route;

[0018] Based on the risk type probability distribution value of each node in each climbing route and the weight of the risk type probability distribution value of each node in each climbing route, calculate the basic risk status indicator of each node in each climbing route according to the weighted fusion formula.

[0019] Furthermore, obtain the training preparation data of the target snow mountain landing user, establish an observation window in unit time, take the training preparation data as the observation object, and extract the travel feature data of the target snow mountain landing user;

[0020] Using time series analysis, mark the travel characteristic parameters of the target snow mountain landing users per unit time in the comprehensive characteristic data of the target snow mountain landing users, and establish a time series set of the travel characteristics of the target snow mountain landing users; the travel characteristic parameters include: stride asymmetry and heart rate change rate.

[0021] Based on the time series set of the travel characteristics of the target snow mountain landing users, train the LSTM long short-term memory network, use the travel characteristic parameters of the target snow mountain landing users per unit time as the input, and use the confidence probability of the fatigue state of the target snow mountain landing users per unit time as the output to obtain a fatigue state evaluation model.

[0022] Furthermore, perform data preprocessing based on the travel state data and environmental data of the real-time snow mountain login users on the route.

[0023] According to the travel state data and environmental state data of the snow mountain landing users on the route after preprocessing, substitute them into the fatigue state evaluation model to obtain the travel state deviation parameters of the real-time snow mountain landing users on the route; the environmental data includes: altitude and slope.

[0024] Use the CUSUM control chart to statistically analyze the travel state deviation parameters of the real-time snow mountain landing users on the route, and calculate the fatigue state index of the corresponding nodes on the route of the real-time snow mountain landing users.

[0025] Furthermore, according to the virtual simulation snow mountain digital twin model, determine several risk types in the corresponding nodes on the route of the real-time snow mountain landing users.

[0026] According to several risk types in the corresponding nodes on the route of the real-time snow mountain landing users, use the K-means clustering function to divide and cluster the environmental state data of the real-time snow mountain landing users on the route to obtain a set of real-time influencing environmental factors for several risk types in the corresponding nodes on the route of the real-time snow mountain landing users.

[0027] Perform normalization processing on the set of real-time influencing environmental factors for several risk types in the corresponding nodes on the route of the real-time snow mountain landing users.

[0028] Furthermore, based on the set of real-time influencing environmental factors for several risk types in the corresponding nodes on the route of the real-time snow mountain landing users, calculate the information entropy of the real-time influencing environmental factors for the target risk type, and determine the weight of the real-time influencing environmental factors.

[0029] Train a GMM Gaussian mixture model according to the set of real-time impact environmental factors of several risk types in the corresponding nodes along the route where the snow mountain landing user is located, with the real-time impact environmental factors as the output and the probabilities of several risk types in the corresponding nodes along the route to which the real-time impact environmental factors belong as the output;

[0030] Based on the weights of the real-time impact environmental factors of several risk types in the corresponding nodes along the route where the snow mountain landing user is located and the probabilities of several risk types in the corresponding nodes along the route to which the real-time impact environmental factors belong, calculate the dynamic risk index of the corresponding nodes along the route where the real-time snow mountain landing user is located.

[0031] Furthermore, based on the fatigue state index of the user along the corresponding node of the route during real-time snow mountain landing, use AR autoregression to estimate the maximum travel ability index of the user during real-time snow mountain landing.

[0032] Furthermore, perform a difference operation based on the fatigue state index of the user along the corresponding node of the route during real-time snow mountain landing and the maximum travel ability index of the user during real-time snow mountain landing to determine the redundancy value of the travel ability index of the user along the corresponding node of the route during real-time snow mountain landing;

[0033] Perform a linear mapping on the redundancy value of the travel ability index and the dynamic risk index of the user along the corresponding node of the route during real-time snow mountain landing to obtain the redundancy value vector of the travel ability index and the dynamic risk index vector of the user along the corresponding node of the route during real-time snow mountain landing;

[0034] According to the Euclidean formula, calculate whether the redundancy value vector of the travel ability index of the user along the corresponding node of the route during real-time snow mountain landing satisfies the dynamic risk index vector of the user along the corresponding node of the route during real-time snow mountain landing.

[0035] Furthermore, a snow mountain landing safety warning system based on an intelligent helmet includes:

[0036] Basic risk module, data collection module, user fatigue state module, basic risk compensation module, safety warning module;

[0037] The basic risk module is used to determine several climbing routes of the snow mountain landing task, collect multi-dimensional risk data of several climbing routes, quantify the basic risk state indicators of each node of each climbing route, and construct a virtual simulation snow mountain digital twin model;

[0038] The data collection module is used to obtain the travel state data and environmental state data of the real-time snow mountain landing user along the route based on the sensors embedded in the intelligent helmet;

[0039] The user fatigue status module is electrically connected to the data acquisition module, and the user fatigue status module is used to evaluate the fatigue status index of the corresponding node of the route where the user is located in the real-time snow mountain landing according to the change trend of the travel status data and environmental status data of the route where the real-time snow mountain landing user is located;

[0040] The basic risk compensation module is electrically connected to the data acquisition module and the basic risk module. The basic risk compensation module is used to substitute the environmental status data of the real-time snow mountain landing user's route into the virtual simulation snow mountain digital twin model, and perform real-time compensation for the basic risk status indicators of each node of each mountaineering route, and quantify the dynamic risk index of the corresponding node under the route where the real-time snow mountain landing user is located;

[0041] The safety warning module is electrically connected to the user fatigue status module and the basic risk compensation module. The safety warning module is used to determine whether the fatigue status index at the corresponding node of the user's route in the real-time snow mountain login meets the dynamic risk index of the corresponding node under the user's route in the real-time snow mountain login. If so, it is determined to continue logging in; if not, it is determined to suspend logging in.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention proposes a multimodal data safety warning smart helmet solution for snow mountain landing. By building a digital twin model of snow mountains, static route risk data (such as terrain, climate, etc.) is integrated, and the smart helmet sensor is used to collect the user's travel status (such as heart rate, cadence) and environmental data (such as wind speed, temperature) in real time, and dynamically evaluate the user's fatigue level and node risk index; the model is used to compensate for the impact of environmental changes on basic risks in real time, and dynamically decide whether to continue climbing. The beneficial effect is to achieve intelligent early warning of mountaineering safety, reduce the risk of human misjudgment through multi-dimensional data linkage, and improve the efficiency of emergency response in extreme environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a flow chart of a safety warning method for snow mountain landing based on a smart helmet;

[0045] Figure 2 This is a framework diagram of a snow mountain landing safety warning system based on a smart helmet; DETAILED DESCRIPTION

[0046] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0047] Reference Figure 1 As shown, a snow mountain landing safety warning method based on a smart helmet comprises:

[0048] Step 1: Determine several climbing routes for the snow mountain landing mission, collect multi-dimensional risk data for several climbing routes, quantify the basic risk status indicators of each node on each climbing route, and construct a virtual simulation snow mountain digital twin model;

[0049] The above Step 1 includes the following contents:

[0050] Step 101: Based on the snow mountain to be landed, determine several climbing routes for the snow mountain landing mission, and obtain multi-dimensional risk data for several climbing routes; the multi-dimensional data of the climbing routes includes: terrain data, meteorological data, and hidden danger distribution data;

[0051] Taking the supply camps on the climbing routes as division nodes, divide the multi-dimensional risk data for several climbing routes to obtain the multi-dimensional risk data of each node on each climbing route;

[0052] Based on the multi-dimensional risk data of each node on each climbing route, perform Z-score standardization preprocessing for numerical data and One-Hot encoding preprocessing for categorical data;

[0053] According to the Transformer architecture, train the modal classifier of the corresponding dimension type according to the multi-dimensional risk data of each node on each climbing route, and generate the risk type logical value of each node on each climbing route;

[0054] Using the Softmax function, with the risk type logical value of each node on each climbing route as the input and the risk type probability distribution value of each node on each climbing route as the output, the method is as follows:

[0055]

[0056] Among them, is the k-th risk type probability distribution value of the j-th node on the i-th climbing route, is the k-th risk type logical value of the j-th node on the i-th climbing route, e is the natural constant, and n is the total number of risk types.

[0057] Step 102: Calculate the contribution of the risk type probability distribution value of each node on each climbing route according to the entropy weight method, and assign weights to the risk type probability distribution value of each node on each climbing route;

[0058] Based on the risk type probability distribution value of each node on each climbing route and the weight of the risk type probability distribution value of each node on each climbing route, calculate the basic risk status indicator of each node on each climbing route according to the weighted fusion formula, and the method is as follows;

[0059]

[0060] Among them, is the basic risk status index of the j-th node of the i-th mountain climbing route, is the weight of the probability distribution value of the k-th risk type of the j-th node of the i-th mountain climbing route.

[0061] When in use, combine the content in Steps 101 to 102:

[0062] As a further content, according to the Transformer architecture, for the multi-dimensional risk data of each node of each mountain climbing route, train the modal classifier of the corresponding dimension type. The modal classifier body is divided into an image classifier and a text classifier. By analyzing the multi-dimensional risk data of each node of the known mountain climbing route for the risk types existing in the route (ice crack distribution, ice fall distribution, avalanche distribution), to achieve quantifying the risk status between each node of the known mountain climbing route and improving the warning accuracy.

[0063] Step 2: Based on the sensors embedded in the intelligent helmet, obtain the travel status data and environmental status data of the user on the real-time snow mountain climbing route;

[0064] Step 3: According to the change trends of the travel status data and environmental status data of the user on the real-time snow mountain climbing route, evaluate the fatigue status index of the corresponding node of the user on the real-time snow mountain climbing route;

[0065] The said Step 3 includes the following content:

[0066] Step 301: Obtain the training preparation data of the target snow mountain climbing user, establish an observation window with a unit time, use the training preparation data as the observation object, and extract the travel feature data of the target snow mountain climbing user;

[0067] Using time series analysis, mark the travel feature parameters of the target snow mountain climbing user in the unit time in the comprehensive feature data of the target snow mountain climbing user, and establish the travel feature time series set of the target snow mountain climbing user; The said travel feature parameters include: stride asymmetry, heart rate change rate;

[0068] Based on the travel feature time series set of the target snow mountain climbing user, train the LSTM long short-term memory network, use the travel feature parameters of the target snow mountain climbing user in the unit time as the input, and use the confidence probability of the fatigue status of the target snow mountain climbing user in the unit time as the output to obtain the fatigue status evaluation model;

[0069] Step 302: Perform data preprocessing based on the travel status data and environmental data of the user on the real-time snow mountain climbing route.

[0070] According to the movement status data and environmental status data of the snow mountain landing user under the preprocessed route, substitute them into the fatigue status evaluation model to obtain the movement status deviation parameter of the snow mountain landing user under the real-time route; the environmental data includes: altitude, slope;

[0071] Use the CUSUM control chart to count the movement status deviation parameters of the snow mountain landing user under the real-time route, and calculate the fatigue status index of the corresponding node on the route of the user during the real-time snow mountain landing.

[0072] When in use, combine the content in steps 301-302:

[0073] As a further content, the traditional CUSUM static threshold has the problem of insufficient adaptability due to physiological baseline drift and environmental interference in high-altitude environments. Therefore, by calculating the mean μ and standard deviation σ of the confidence probability of the fatigue status in the real-time rolling window, and introducing blood oxygen saturation (SpO2) and slope as compensation factors, the threshold multiple is dynamically adjusted to make the threshold adaptively relax as the altitude increases, blood oxygen decreases, or slope increases, so as to maintain a low false alarm rate while avoiding missing the true fatigue status. And because the computing power of the smart helmet as an edge device is limited, therefore, the SLTM long short-term memory network is pre-trained based on the training preparation data of the historical target snow mountain landing user and deployed in the cloud. However, due to the lack of snow mountain environmental factors during the training process, during subsequent real-time snow mountain landings, by using real-time environmental data, the fatigue status evaluation model in the cloud is compensated to achieve accurate evaluation of the movement status of the snow mountain landing user.

[0074] Step Four: Substitute the environmental status data of the real-time snow mountain landing user under the route into the virtual simulation snow mountain digital twin model, and perform real-time compensation for the basic risk status indicators of each node on each climbing route, and quantify the dynamic risk index of the corresponding node under the route of the real-time snow mountain landing user.

[0075] The content of Step Four includes the following:

[0076] Step 401: According to the virtual simulation snow mountain digital twin model, determine several risk types in the corresponding nodes under the route of the real-time snow mountain landing user.

[0077] According to several risk types in the corresponding nodes under the route of the real-time snow mountain landing user, use the K-means clustering function to partition and cluster the environmental status data of the real-time snow mountain landing user under the route, and obtain the set of real-time impact environmental factors of several risk types in the corresponding nodes under the route of the real-time snow mountain landing user.

[0078] Normalize the set of real-time impact environmental factors for several risk types in the corresponding nodes along the route of the real-time snow mountain landing user;

[0079] Step 402: Calculate the information entropy of the real-time impact environmental factors for the target risk type based on the set of real-time impact environmental factors for several risk types in the corresponding nodes along the route of the real-time snow mountain landing user, and determine the weights of the real-time impact environmental factors;

[0080] Train a GMM Gaussian mixture model according to the set of real-time impact environmental factors for several risk types in the corresponding nodes along the route of the snow mountain landing user, with the real-time impact environmental factors as the output and the probabilities of several risk types in the corresponding nodes along the route to which the real-time impact environmental factors belong as the output;

[0081] Based on the weights of the real-time impact environmental factors for several risk types in the corresponding nodes along the route of the snow mountain landing user and the probabilities of several risk types in the corresponding nodes along the route to which the real-time impact environmental factors belong, calculate the dynamic risk index of the corresponding nodes along the route of the real-time snow mountain landing user, and the method is as follows:

[0082]

[0083] Among them, is the dynamic risk index of the j-th node on the i-th route of the real-time snow mountain landing user, is the real-time impact environmental factor The probability of belonging to several risk types R in the corresponding nodes along the route, k of, is the weight of the real-time impact environmental factors of the j-th node on the i-th route of the real-time snow mountain landing user;

[0084] When in use, combine the content in 401 to 402:

[0085] As further content, since the existing virtual simulation snow mountain digital twin models are all obtained by fitting historical data, and the actual environment of snow mountain landing is complex and cannot be predicted by common weather models, therefore, by collecting environmental data in real time through an intelligent helmet and dynamically compensating the existing virtual simulation snow mountain digital twin models, the real-time performance and reliability of risk monitoring in the snow mountain scene can be significantly improved, providing a quantitative basis for safe climbing.

[0086] Step Five: Judge whether the fatigue state index under the corresponding node of the route where the user is located in the real-time snow mountain landing meets the dynamic risk index of the corresponding node along the route of the real-time snow mountain landing user. If so, it is determined to continue logging in. If not, it is determined to suspend logging in;

[0087] The said Step Five includes the following content:

[0088] Step 501: Based on the fatigue state index of the corresponding node on the user's route during real-time snow mountain landing, use autoregressive (AR) to estimate the maximum travel ability index of the user during real-time snow mountain landing;

[0089] Step 502: Perform a difference operation based on the fatigue state index of the corresponding node on the user's route during real-time snow mountain landing and the maximum travel ability index of the user during real-time snow mountain landing to determine the redundancy value of the travel ability index of the corresponding node on the user's route during real-time snow mountain landing;

[0090] Perform a linear mapping on the redundancy value of the travel ability index of the corresponding node on the user's route during real-time snow mountain landing and the dynamic risk index to obtain the redundancy value vector of the travel ability index and the dynamic risk index vector of the corresponding node on the user's route during real-time snow mountain landing;

[0091] According to the Euclidean formula, calculate whether the redundancy value vector of the travel ability index of the corresponding node on the user's route during real-time snow mountain landing meets the dynamic risk index vector of the corresponding node on the user's route during real-time snow mountain landing, in the following manner:

[0092]

[0093] where, is the spatial distance between the redundancy value vector of the travel ability index of the corresponding node on the user's route during real-time snow mountain landing and the dynamic risk index vector of the corresponding node on the user's route during real-time snow mountain landing, is the redundancy value vector of the travel ability index of the j-th node on the i-th route of the user during real-time snow mountain landing, is the dynamic risk index vector of the j-th node on the i-th route of the user during real-time snow mountain landing, and m is the total number of nodes under the route.

[0094] Refer to Figure 2 As shown, a snow mountain landing safety warning system based on an intelligent helmet includes:

[0095] Basic risk module, data acquisition module, user fatigue state module, basic risk compensation module, safety warning module;

[0096] The basic risk module is used to determine several climbing routes for the snow mountain landing task, collect multi-dimensional risk data in several climbing routes, quantify the basic risk state index of each node in each climbing route, and construct a virtual simulation snow mountain digital twin model;

[0097] The data acquisition module is used to obtain the travel state data and environmental state data of the user on the current route during real-time snow mountain landing based on the sensors embedded in the intelligent helmet;

[0098] The user fatigue state module is electrically connected to the data acquisition module. The user fatigue state module is used to evaluate the fatigue state index of the corresponding node on the route where the user is located during the real-time snow mountain landing according to the change trend of the progress state data and the environmental state data under the route where the real-time snow mountain landing user is located.

[0099] The basic risk compensation module is electrically connected to the data acquisition module and the basic risk module. The basic risk compensation module is used to substitute the environmental state data under the route where the real-time snow mountain landing user is located into the virtual simulation snow mountain digital twin model, and perform real-time compensation for the basic risk state indexes of each node of each climbing route, and quantify the dynamic risk index of the corresponding node under the route where the real-time snow mountain landing user is located.

[0100] The safety warning module is electrically connected to the user fatigue state module and the basic risk compensation module. The safety warning module is used to judge whether the fatigue state index of the corresponding node on the route where the user is located during the real-time snow mountain landing meets the dynamic risk index of the corresponding node under the route where the real-time snow mountain landing user is located. If so, it is determined to continue logging in. If not, it is determined to suspend logging in.

[0101] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. A safety warning method for snow mountain landing based on an intelligent helmet, characterized in that, Including: S1. Determine several climbing routes for the snow mountain landing mission, collect multi-dimensional risk data for several climbing routes, quantify the basic risk status indicators of each node on each climbing route, and construct a virtual simulation snow mountain digital twin model; S2. Based on the sensors embedded in the intelligent helmet, obtain the movement status data and environmental status data of the real-time snow mountain landing user on the current route; S3. According to the change trend of the movement status data and environmental status data of the real-time snow mountain landing user on the current route, evaluate the fatigue status indicator of the corresponding node on the route where the user is located during the real-time snow mountain landing; S4. Substitute the environmental status data of the real-time snow mountain landing user on the current route into the virtual simulation snow mountain digital twin model, perform real-time compensation for the basic risk status indicators of each node on each climbing route, and quantify the dynamic risk index of the corresponding node on the route where the real-time snow mountain landing user is located; S4. Determine whether the fatigue status indicator of the corresponding node on the route where the user is located during the real-time snow mountain landing meets the dynamic risk index of the corresponding node on the route where the real-time snow mountain landing user is located. If so, it is determined to continue logging in. If not, it is determined to suspend logging in.

2. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 1, wherein The above S1 includes: Based on the snow mountain to be landed, determine several climbing routes for the snow mountain landing mission, and obtain multi-dimensional risk data for several climbing routes; the multi-dimensional data of the climbing route includes: terrain data, meteorological data, and hidden danger distribution data; Taking the supply camps in the climbing route as the division nodes, divide the multi-dimensional risk data for several climbing routes to obtain the multi-dimensional risk data of each node on each climbing route; Based on the multi-dimensional risk data of each node on each climbing route, perform Z-score standardization preprocessing for numerical data and One-Hot encoding preprocessing for categorical data; According to the Transformer architecture, train the modal classifier of the corresponding dimension type according to the multi-dimensional risk data of each node on each climbing route, and generate the risk type logical value of each node on each climbing route; Use the Softmax function, with the risk type logical value of each node on each climbing route as the input and the risk type probability distribution value of each node on each climbing route as the output.

3. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 2, characterized in that The above S1 further includes: According to the entropy weight method, calculate the contribution of the risk type probability distribution value of each node on each climbing route, and assign weights to the risk type probability distribution value of each node on each climbing route; Based on the risk type probability distribution value of each node on each climbing route and the weight of the risk type probability distribution value of each node on each climbing route, calculate the basic risk status indicator of each node on each climbing route according to the weighted fusion formula.

4. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 1, characterized in that The above S3 includes: Obtain the training preparation data of the target snow mountain landing user, establish an observation window per unit time, use the training preparation data as the observation object, and extract the movement characteristic data of the target snow mountain landing user; Using time series analysis, mark the travel feature parameters of the target snow mountain landing users per unit time in the comprehensive feature data of the target snow mountain landing users, and establish a time series set of the travel features of the target snow mountain landing users; the travel feature parameters include: stride asymmetry and heart rate change rate. Based on the time series set of the travel features of the target snow mountain landing users, train the LSTM long short-term memory network, using the travel feature parameters of the target snow mountain landing users per unit time as the input and the confidence probability of the fatigue state of the target snow mountain landing users per unit time as the output to obtain a fatigue state evaluation model.

5. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 4, wherein The S3 also includes: Perform data preprocessing based on the travel state data and environmental data of the real-time snow mountain landing users on the route. According to the travel state data and environmental state data of the snow mountain landing users on the route after preprocessing, substitute them into the fatigue state evaluation model to obtain the travel state deviation parameter of the real-time snow mountain landing users on the route; the environmental data includes: altitude and slope. Use the CUSUM control chart to statistically analyze the travel state deviation parameters of the real-time snow mountain landing users on the route, and calculate the fatigue state index of the corresponding nodes of the users on the route during the real-time snow mountain landing.

6. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 2, wherein The S4 includes: According to the virtual simulation snow mountain digital twin model, determine several risk types in the corresponding nodes of the real-time snow mountain landing users on the route. According to several risk types in the corresponding nodes of the real-time snow mountain landing users on the route, use the K-means clustering function to partition and cluster the environmental state data of the real-time snow mountain landing users on the route to obtain a set of real-time influencing environmental factors for several risk types in the corresponding nodes of the real-time snow mountain landing users on the route. Perform normalization processing on the set of real-time influencing environmental factors for several risk types in the corresponding nodes of the real-time snow mountain landing users on the route.

7. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 6, wherein S4 also includes: Based on the set of real-time influencing environmental factors for several risk types in the corresponding nodes of the real-time snow mountain landing users on the route, calculate the information entropy of the real-time influencing environmental factors for the target risk type to determine the weight of the real-time influencing environmental factors. According to the set of real-time influencing environmental factors for several risk types in the corresponding nodes of the real-time snow mountain landing users on the route, train the GMM Gaussian mixture model, with the real-time influencing environmental factors as the output and the probability of several risk types in the corresponding nodes under the route to which the real-time influencing environmental factors belong as the output. Based on the weight of the real-time influencing environmental factors for several risk types in the corresponding nodes of the snow mountain landing users on the route and the probability of several risk types in the corresponding nodes under the route to which the real-time influencing environmental factors belong, calculate the dynamic risk index of the corresponding nodes of the real-time snow mountain landing users on the route.

8. The method for snow mountain landing safety warning based on an intelligent helmet according to claim 7, wherein The S5 includes: Based on the fatigue state index of the corresponding nodes of the users during the real-time snow mountain landing, use AR autoregression to estimate the maximum travel ability index of the users during the real-time snow mountain landing.

9. A snow mountain landing safety warning method based on an intelligent helmet according to claim 4, characterized in that The S5 also includes: Perform a difference operation based on the fatigue state index at the corresponding node of the user's route in the real-time snow mountain landing and the maximum travel ability index of the user in the real-time snow mountain landing to determine the redundancy value of the travel ability index at the corresponding node of the user's route in the real-time snow mountain landing; Perform a linear mapping on the redundancy value of the travel ability index at the corresponding node of the user's route in the real-time snow mountain landing and the dynamic risk index to obtain the redundancy value vector of the travel ability index and the dynamic risk index vector at the corresponding node of the user's route in the real-time snow mountain landing; According to the Euclidean formula, calculate whether the redundancy value vector of the travel ability index at the corresponding node of the user's route in the real-time snow mountain landing satisfies the dynamic risk index vector at the corresponding node of the user's route in the real-time snow mountain landing.

10. A snow mountain landing safety warning system based on an intelligent helmet, comprising: A basic risk module, a data collection module, a user fatigue state module, a basic risk compensation module, and a safety warning module; The basic risk module is used to determine several climbing routes for the snow mountain landing task, collect multi-dimensional risk data in several climbing routes, quantify the basic risk state index of each node of each climbing route, and construct a virtual simulation snow mountain digital twin model; The data collection module is used to obtain the travel state data and environmental state data of the user in the real-time snow mountain landing based on the sensors embedded in the intelligent helmet; The user fatigue state module is electrically connected to the data collection module. The user fatigue state module is used to evaluate the fatigue state index at the corresponding node of the user's route in the real-time snow mountain landing according to the change trend of the travel state data and environmental state data of the user in the real-time snow mountain landing; The basic risk compensation module is electrically connected to the data collection module and the basic risk module. The basic risk compensation module is used to substitute the environmental state data of the user in the real-time snow mountain landing into the virtual simulation snow mountain digital twin model, perform real-time compensation on the basic risk state index of each node of each climbing route, and quantify the dynamic risk index of the corresponding node of the user in the real-time snow mountain landing; The safety warning module is electrically connected to the user fatigue state module and the basic risk compensation module. The safety warning module is used to judge whether the fatigue state index at the corresponding node of the user's route in the real-time snow mountain landing satisfies the dynamic risk index of the corresponding node of the user in the real-time snow mountain landing. If so, it is determined to continue logging in. If not, it is determined to suspend logging in.

Images