A snow avalanche monitoring and vehicle safety method for mountain road snow removal operation

Abstract

The application discloses a snow avalanche monitoring and vehicle safety method for snow removal operation of mountainous area highways, relates to the technical field of geological disaster monitoring and intelligent traffic safety, and has the following steps: a system receives real-time monitoring data sent by an environment monitoring device and vehicle position data sent by a snow removal vehicle; the system calculates a snow avalanche occurrence probability in a future period of time by using a preset prediction model according to the real-time monitoring data; when the snow avalanche occurrence probability exceeds a set value, the system generates a danger zone with the monitoring device as the center; the system calculates whether the snow removal vehicle is located in the danger zone in real time; if the snow removal vehicle is located in the danger zone, the system issues an alarm on a visual large screen and sends a withdrawal instruction to the snow removal vehicle. The application considers both sudden and gradual disaster monitoring, solves the problem that traditional two-dimensional monitoring cannot reflect the risk of terrain elevation difference, and realizes automatic and accurate safety of end cloud cooperation.

Description

Technical Field

[0001] This invention relates to the field of geological disaster monitoring and intelligent transportation safety technology, and in particular to an avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads. Background Technology

[0002] High-altitude mountain roads are vital transportation arteries connecting remote areas, and are highly susceptible to continuous snowfall in winter, leading to snow accumulation and icing. To ensure road accessibility, mechanized snow removal has become the primary means of routine maintenance. Snow removal vehicle drivers need to work for extended periods in complex mountainous terrain and harsh weather conditions, and their attention is primarily focused on clearing the road surface, making it difficult to monitor the snow accumulation on the slopes above the road. When the snow accumulation on the slopes becomes structurally unstable due to gravity, wind, or temperature changes, it can easily trigger avalanches, posing a serious threat to the safety of snow removal vehicles and personnel.

[0003] Existing methods for monitoring natural disasters on mountain roads largely rely on manual patrols or basic meteorological station monitoring. Manual patrols are inefficient and have blind spots, failing to meet the need for real-time, 24 / 7 monitoring. While traditional meteorological monitoring stations can collect basic data such as temperature, wind speed, and snow depth, they often lack the ability to detect microscopic changes in the snow layer, making it difficult to capture subtle displacements or structural slippage characteristics of snow before avalanches. Most systems rely solely on single, fixed thresholds for passive alarms, failing to fully utilize long-term historical time-series data for trend prediction, resulting in significant lag in early warning mechanisms and hindering the ability to buy valuable evacuation time for workers before disasters occur.

[0004] Existing monitoring systems primarily present information in the form of two-dimensional maps or data tables. Due to the significant elevation differences and winding roads in mountainous terrain, two-dimensional maps cannot accurately represent the impact of terrain undulations on the scope of disasters, making it difficult for monitoring personnel to intuitively determine the three-dimensional spatial relationship between hazards and vehicles. Furthermore, environmental monitoring systems and vehicle dispatching systems often exist as information silos, operating independently on different platforms. Warning information needs to be manually interpreted and then transmitted to drivers via walkie-talkies or other voice devices. This non-automated transmission chain prolongs information flow time and fails to provide drivers with intuitive visual evacuation route guidance, easily leading to vehicles missing the best escape route during emergency avoidance due to panic or poor visibility. Summary of the Invention

[0005] In response to the problems of outdated monitoring methods, lack of micro-displacement sensing capabilities, and poor disaster early warning information transmission links in existing technologies, this invention provides an avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for avalanche monitoring and vehicle avoidance in snow removal operations on mountain roads, comprising the following steps:

[0007] S1: The system receives real-time monitoring data sent by environmental monitoring equipment and vehicle location data sent by snowplows;

[0008] S2: The system calculates the probability of avalanche occurrence in the future using a preset prediction model based on real-time monitoring data;

[0009] S3: When the probability of an avalanche exceeds the set value, the system generates a danger zone centered on the monitoring equipment;

[0010] S4: The system calculates in real time whether the snowplow is located within the danger zone;

[0011] S5: If the snowplow is located in a dangerous area, the system will issue an alarm on the large visual screen and send an evacuation command to the snowplow.

[0012] Furthermore, the data received in S1 is as follows: by subscribing to data packets of snow offset, temperature, humidity, snow depth and wind speed published by the environmental monitoring equipment through the MQTT message queue protocol; and by subscribing to latitude and longitude coordinate data published by the snow removal vehicle terminal through the MQTT protocol.

[0013] Furthermore, before S2, there is a dual-trigger judgment step: the system presets an absolute safety threshold for snow offset; it judges whether the real-time monitored snow offset directly exceeds the preset displacement safety threshold; if it exceeds, it determines that the snow layer structure is unstable, skips the prediction model calculation, and directly executes S3 to generate the danger zone; if it does not exceed, it executes S2 and uses the prediction model to calculate the probability of avalanche occurrence.

[0014] Preferably, the prediction model in S2 uses an LSTM neural network. Its specific construction and processing include: removing outliers from the received snow offset, snow depth, temperature, humidity, and wind speed data, and mapping the data to the [0,1] interval using the Min-Max normalization method; setting a time step and using a sliding window algorithm to convert the normalized data into a multi-dimensional time series feature vector as the model input; inputting the feature vector into the LSTM network, extracting time-dependent features using forgetting and input gate mechanisms, and mapping the high-dimensional features to a one-dimensional output through a fully connected layer; and processing the output of the fully connected layer using the Sigmoid activation function to obtain avalanche occurrence probability values ​​between 0 and 1.

[0015] Furthermore, the specific method for generating the danger zone in S3 is as follows: when the probability of avalanche occurrence exceeds a set value, the three-dimensional geographic coordinates of the monitoring device are obtained, and the local terrain slope value is calculated based on the digital elevation model data around the coordinates; a reference radius and slope influence coefficient are set, and a positive correlation strategy is used to calculate the dynamic warning radius R, wherein the larger the terrain slope value, the larger the dynamic warning radius R; a cylindrical warning area perpendicular to the ground plane is constructed with the geographic coordinates of the monitoring device as the center of the base circle and R as the radius as the danger zone.

[0016] Furthermore, the specific calculation process of S4 is as follows: real-time acquisition of the current three-dimensional coordinates of the snowplow; calculation of the spatial straight-line distance between the snowplow coordinates and the monitoring equipment coordinates; determination of whether the straight-line distance is less than the dynamic warning radius R; if it is less than R, it is determined that the snowplow is located in the danger zone.

[0017] Furthermore, the specific execution process of alarm and instruction sending in step S5 includes: rendering the danger zone as a semi-transparent red three-dimensional light column on the three-dimensional electronic map and marking the snowplow model located in the danger zone as a bright flashing state; automatically generating a recommended path from the current position of the snowplow to the safe area; displaying the recommended path on the three-dimensional electronic map and simultaneously encapsulating it into a navigation instruction data packet and sending it to the snowplow's on-board terminal via the network.

[0018] As a preferred embodiment, the specific construction and updating method of the three-dimensional electronic map is as follows: loading the digital elevation model and satellite image data of the target mountain area as a static terrain base; reading the real-time latitude and longitude coordinates of the environmental monitoring equipment and snowplows, and mapping them as dynamic three-dimensional points on the terrain base; parsing the alarm status field in the real-time monitoring data, and automatically rendering the corresponding point model as a red alarm form when the status is abnormal.

[0019] In summary, the beneficial effects of this invention are as follows:

[0020] 1. Compared with existing technologies, this invention achieves substantial breakthroughs in monitoring dimensions and early warning mechanisms. By innovatively introducing snow cover displacement as a core monitoring indicator, the system can directly capture the instability characteristics of the snow layer structure, overcoming the lag of traditional passive monitoring that relies solely on snow depth. Combined with a dual triggering mechanism, it can immediately trigger an alarm when sudden snow cover displacement occurs, and can also utilize an LSTM deep learning model to mine the temporal dependency features of long-term data for trend prediction. This achieves a leap from passive post-disaster alarms to proactive pre-disaster early warnings, significantly improving the timeliness and accuracy of disaster monitoring in complex high-altitude mountainous environments.

[0021] 2. This invention offers a more scientific and precise approach to spatial risk assessment. Addressing the characteristics of mountainous terrain with significant elevation differences and steep slopes, this invention abandons the traditional method of drawing circles with a fixed radius on a plane. Instead, it utilizes a Digital Elevation Model (DEM) to calculate local terrain slopes in real time and employs a positive correlation strategy to dynamically calculate the warning radius, constructing a cylindrical warning zone perpendicular to the ground plane. This three-dimensional spatial modeling method based on real terrain data fully considers the physical laws governing the expanded avalanche impact range in steep terrain, effectively solving the technical challenge of two-dimensional maps failing to intuitively reflect elevation risk.

[0022] 3. This invention achieves efficient automated hazard avoidance scheduling through edge-cloud collaboration. The system utilizes the MQTT protocol to establish an information link between the environmental monitoring terminal and the vehicle-mounted operation terminal. Once a snowplow is determined to have entered a dangerous area, the system not only issues a visual alarm on the command center's large screen but also automatically generates a recommended evacuation route and encapsulates it into a command packet that is directly sent to the vehicle-mounted terminal. This automated closed-loop command transmission mechanism maximizes the safety of snowplow operation personnel. Attached Figure Description

[0023] Figure 1 The flowchart shown is a method for avalanche monitoring and vehicle avoidance in snow removal operations on mountain roads according to the present invention.

[0024] Figure 2 The diagram shows the structure of the LSTM neural network prediction model used in the avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to the present invention. Detailed Implementation

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] Please see Figure 1 This invention provides an embodiment: an avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads, the steps of which are as follows:

[0027] S1: The system receives real-time monitoring data from environmental monitoring equipment and vehicle location data from snowplows. The system subscribes to snow cover offset data published by roadside sensing terminals via the MQTT protocol. Snow depth ,temperature ,humidity and wind speed The five-dimensional environmental vector is constructed; at the same time, the real-time latitude and longitude coordinates of the snowplow are subscribed to by the vehicle terminal.

[0028] S2: The system calculates the probability of avalanches occurring within a future period using a preset prediction model based on real-time monitoring data. To balance the instantaneous response to sudden avalanches with the trend prediction of gradual avalanches, this embodiment employs a dual-trigger logic of "hard threshold" and "soft prediction" in parallel before model prediction. The system presets an absolute safety threshold for snow accumulation offset. For 20 The system determines the currently monitored snow cover offset in real time. Is it greater than If the value exceeds this threshold, the snow layer structure is deemed to have experienced substantial physical instability. In this case, no model prediction is required; subsequent calculations are skipped, an alarm is forcibly triggered, and the process proceeds to step S3 to generate the danger zone. If the value does not exceed this threshold, data preprocessing and LSTM model calculations are performed.

[0029] When entering the LSTM model computation process, the received five-dimensional data is first mapped to the [0,1] interval using the Min-Max normalization method. The normalization formula is as follows:

[0030]

[0031] This is the raw monitoring data. and These are the historical minimum and maximum values ​​for this type of data, respectively. These are the standardized input values.

[0032] Please see Figure 2 The LSTM neural network prediction model used in this embodiment includes forget gate, input gate, and output gate mechanisms. The time step is set to... The five-dimensional feature vectors obtained after the Min-Max standardization process are then... As the input vector at the current moment The hidden state at the previous moment was The cell state at the previous moment was .

[0033] The specific calculation process is as follows: First, calculate the forget gate. The formula for determining which information to discard from the cell's state at the previous moment is:

[0034]

[0035] Next, calculate the input gate. The formula determines which new information will be updated to the cell state at the current moment:

[0036]

[0037]

[0038] Then update the cell status. The formula is:

[0039]

[0040] Finally, calculate the output gate. and hidden state The formula is:

[0041]

[0042]

[0043] In the above formula, It is the Sigmoid activation function. The hyperbolic tangent activation function is used. , , , These are the weight matrices for the forget gate, input gate, cell state update, and output gate, respectively. , , , These are the corresponding bias vectors. This indicates the current state of the candidate cells, i.e., after [the current state]. New information candidate values ​​after activation This represents concatenating the hidden state from the previous time step with the current input vector. The hidden state output by the LSTM is then used. The avalanche probability is then calculated using the Sigmoid function after being passed into a fully connected layer. The specific formula is as follows:

[0044]

[0045] In the formula This is the weight matrix of the fully connected layer, used to map the high-dimensional hidden state to a one-dimensional output space; This is the bias term for the fully connected layer. If the calculated probability... If the alarm threshold of 0.8 set by the system is exceeded, it is determined that there is a high risk and proceeds to step S3.

[0046] S3: When the probability of an avalanche exceeds a set value or a hard threshold alarm is triggered, the system generates a danger zone centered on the monitoring device. Considering the characteristics of mountainous terrain with large elevation differences and steep slopes, this embodiment uses a "positive correlation strategy" to dynamically calculate the warning radius. The specific calculation process is as follows: First, the three-dimensional geographic coordinates of the monitoring device are obtained, and the coordinates and surrounding areas are read. The local terrain slope value at a given location is calculated by differential calculation using digital elevation model (DEM) data of the neighborhood within the grid. The dynamic early warning radius is then calculated using the following formula. :

[0047]

[0048] In the formula, The preset reference radius is 500 meters in this embodiment; The slope influence coefficient is taken as 0.8 in this embodiment. As can be seen from the formula, the terrain slope value... The larger the value, the greater the calculated dynamic warning radius. The larger the area, the more likely it is to create a cylindrical warning zone perpendicular to the ground plane to cover a wider area affected by avalanches in steep terrain.

[0049] S4: The system calculates in real time whether the snowplow is located within the danger zone. Specifically, the system calculates in real time the straight-line distance between the three-dimensional coordinates of the snowplow and the coordinates of the monitoring equipment. The calculation formula is:

[0050]

[0051] In the formula, The straight-line distance between the vehicle and the monitoring point; The current real-time 3D coordinates of the snowplow; The fixed three-dimensional coordinates of the environmental monitoring equipment. If the calculated distance... Less than the dynamic warning radius If so, the vehicle is determined to be located in a dangerous area.

[0052] S5: If the snowplow is located in a danger zone, the system needs to issue an alarm and provide evacuation guidance. Specifically, on the 3D electronic map in the command center, the aforementioned cylindrical danger zone will be rendered as a semi-transparent red 3D light pillar, and the snowplow model within the pillar will be marked as highlighted and flashing. Simultaneously, the system will automatically calculate a route from the snowplow's current location to the nearest safe area (i.e., a distance greater than...) based on road network data. The shortest recommended path for the region is encapsulated into a navigation instruction data packet and sent to the vehicle terminal via the MQTT protocol for the driver's reference and risk avoidance.

[0053] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for avalanche monitoring and vehicle avoidance in snow removal operations on mountain roads, characterized in that, The steps are as follows: S1: The system receives real-time monitoring data sent by environmental monitoring equipment and vehicle location data sent by snowplows; S2: The system calculates the probability of avalanche occurrence in the future using a preset prediction model based on real-time monitoring data; S3: When the probability of an avalanche exceeds the set value, the system generates a danger zone centered on the monitoring equipment; S4: The system calculates in real time whether the snowplow is located within the danger zone; S5: If the snowplow is located in a dangerous area, the system will issue an alarm on the large visual screen and send an evacuation command to the snowplow.

2. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 1, characterized in that: The method of receiving data in S1 is as follows: subscribing to data packets of snow offset, temperature, humidity, snow depth and wind speed published by the environmental monitoring equipment through the MQTT message queue protocol; and simultaneously subscribing to latitude and longitude coordinate data published by the snow removal vehicle terminal through the MQTT protocol.

3. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 1, characterized in that: The prediction model in S2 uses an LSTM neural network. Its specific construction and processing include: removing outliers from the received snow offset, snow depth, temperature, humidity, and wind speed data, and mapping the data to the [0,1] interval using the Min-Max normalization method; setting a time step and using a sliding window algorithm to convert the normalized data into a multi-dimensional time series feature vector as the input to the model; inputting the feature vector into the LSTM network, extracting time-dependent features using forget gate and input gate mechanisms, and mapping the high-dimensional features to a one-dimensional output through a fully connected layer; and using the Sigmoid activation function to process the output of the fully connected layer to obtain an avalanche occurrence probability value between 0 and 1.

4. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 1, characterized in that: The specific method for generating the danger zone in S3 is as follows: when the probability of avalanche occurrence exceeds a set value, the three-dimensional geographic coordinates of the monitoring device are obtained, and the local terrain slope value is calculated based on the digital elevation model data around the coordinates; a reference radius and slope influence coefficient are set, and a positive correlation strategy is used to calculate the dynamic warning radius R, wherein the larger the terrain slope value, the larger the dynamic warning radius R; a cylindrical warning area perpendicular to the ground plane is constructed with the geographic coordinates of the monitoring device as the center of the base circle and R as the radius as the danger zone.

5. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 4, characterized in that: The specific calculation process of S4 is as follows: real-time acquisition of the current three-dimensional coordinates of the snowplow; calculation of the spatial straight-line distance between the snowplow coordinates and the monitoring equipment coordinates; determination of whether the straight-line distance is less than the dynamic warning radius R; if it is less than R, it is determined that the snowplow is located in the danger zone.

6. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 1, characterized in that: The specific execution process of alarm and instruction sending in step S5 includes: rendering the danger zone as a semi-transparent red three-dimensional light column on the three-dimensional electronic map and marking the snowplow model located in the danger zone as a bright flashing state; automatically generating a recommended path from the current position of the snowplow to the safe area; displaying the recommended path on the three-dimensional electronic map and simultaneously encapsulating it into a navigation instruction data packet and sending it to the snowplow's on-board terminal via the network.

7. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 1, characterized in that: Before S2, there is also a dual-trigger judgment step: the system presets an absolute safety threshold for snow offset; and determines whether the real-time monitored snow offset directly exceeds the preset displacement safety threshold. If the threshold is exceeded, the snow layer structure is deemed unstable, the prediction model calculation is skipped, and S3 is executed directly to generate the danger zone; if the threshold is not exceeded, S2 is executed to calculate the probability of avalanche occurrence using the prediction model.

8. The avalanche monitoring and vehicle avoidance method for snow removal operations on mountain roads according to claim 6, characterized in that: The specific construction and updating method of the three-dimensional electronic map is as follows: load the digital elevation model and satellite image data of the target mountain area as a static terrain base; read the real-time latitude and longitude coordinates of environmental monitoring equipment and snowplows, and map them as dynamic three-dimensional points on the terrain base; parse the alarm status field in the real-time monitoring data, and automatically render the corresponding point model as a red alarm form when the status is abnormal.

Images