Accumulated snow state monitoring system based on wireless sensing technology and cloud service

Abstract

The invention discloses an accumulated snow state monitoring system based on a wireless sensing technology and cloud service, and belongs to the technical field of snow disaster monitoring, and the system comprises a node deployment module which is used for recording node information; the node activation module evaluates whether a node is activated or not; the structure identification module is used for receiving data collected by the activation node and identifying and judging the snow layer structure; the node scheduling module is used for judging the subsequent working state of the node; and the information aggregation module is used for integrating the information collected by the activation node and marking a dangerous area. According to the accumulated snow state monitoring system based on the wireless sensing technology and the cloud service, node sampling frequency, data return behaviors and communication paths are dynamically regulated and controlled through a monitoring mechanism for sensing the state of a multi-layer structure in accumulated snow and an energy state sensing model in combination with the change trend of snow conditions; the operation period of the equipment is effectively prolonged, the maintainability in a severe environment is improved, and priority allocation can be carried out on the monitoring resources based on the identified snow layer state change condition.

Description

Technical Field

[0001] The present invention belongs to the technical field of snow disaster monitoring, and in particular relates to a snow status monitoring system based on wireless sensing technology and cloud services. Background Art

[0002] With the increasing frequency of extreme weather events, snow monitoring has become increasingly important in areas such as public safety, traffic management, power facility operation and maintenance, and disaster warning in mountainous areas. Currently, many regions, particularly those at high altitudes or in remote areas, rely on snow sensor networks to monitor snowfall in real time and prevent risks such as avalanches, icy floods, road closures, and power outages. Existing snow monitoring systems primarily rely on sensors for physical quantities such as snow depth, temperature, and humidity. These sensors often transmit collected data wirelessly to cloud platforms for data visualization and early warning analysis. Furthermore, the recent development of cloud services, low-power IoT devices, and edge computing technologies has also provided technical support for snow monitoring, gradually forming a monitoring system based on "front-end perception + back-end analysis."

[0003] However, current technologies primarily rely on "point-based measurements," indirectly inferring regional snow cover conditions by measuring variables such as snow depth, temperature, and humidity at a specific location. These technologies lack the ability to effectively identify the internal structure of the snowpack (e.g., stratification, cavities, and ice slabs). This makes it impossible for systems to determine whether the snowpack contains potentially unstable structures, severely limiting their applicability in critical scenarios such as avalanche warning and snow pressure assessment. Furthermore, the complexity of the snowy environment makes sensor nodes susceptible to snow cover or damage, limiting communication quality. Furthermore, most systems rely on battery power, significantly impacting energy consumption during long-term operation. Existing systems generally employ periodic sampling and a unified reporting mechanism, failing to intelligently schedule data based on node status, remaining battery power, and changing snow conditions. This results in unnecessary data redundancy and energy waste. Furthermore, due to the significant differences in snowpack evolution across different regions, existing systems lack the ability to prioritize "potential risk points" in data collection and scheduling, leading to delayed responses to sudden risks. In addition, existing systems generally use simple rules or threshold judgments in sensor data processing, and fail to fully utilize structured modeling methods to conduct in-depth analysis and prediction of snow conditions, resulting in the value of data not being effectively mined.

[0004] To this end, we propose a snow status monitoring system based on wireless sensing technology and cloud services to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems in the existing technology of lack of snow layer structure recognition and inability to intelligently schedule, and to propose a snow status monitoring system based on wireless sensing technology and cloud services.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A snow status monitoring system based on wireless sensing technology and cloud services, including:

[0008] A node deployment module, wherein the node deployment unit is used to record basic node information for each node and calculate a risk sensitivity score. The basic node information includes node location information, node terrain conditions, node energy status, and node slope conditions. The risk sensitivity score is obtained by weighted calculation of the basic node information;

[0009] A node activation module is used to determine whether to activate a node and evaluate the node using a node activation score calculated by weighting the node's basic information and risk sensitivity score. Nodes that reach the activation threshold are activated, and after activation, acoustic wave sampling and information recording are performed to obtain acoustic wave reflection data, node temperature, and snow depth.

[0010] A structure recognition module is used to model and identify characteristic structure vectors and output a structural risk score. The characteristic structure vectors include acoustic reflection data, node basic information, node temperature, and snow depth. The structural risk score is calculated using a structure mapping model composed of two layers of linear transformation and activation function.

[0011] A node scheduling module is used to determine whether each node continues sampling or enters sleep mode during the current scheduling cycle. The determination is made by comparing the node scheduling score with the scheduling threshold. The node scheduling score is obtained by weighted calculation of the structural risk score, node energy status, node location information, and temperature change rate input into the scheduling score function. The temperature change rate is calculated based on the node temperature.

[0012] An information aggregation module is used to aggregate the input data of all activated nodes to obtain a regional structural risk map. The input data includes the spatial coordinates of the nodes, the structural risk score, and the node scheduling status. The regional structural risk map is calculated using a spatial weighted interpolation algorithm and then written into a two-dimensional grid or vector map. A structural risk threshold is set. When the structural risk score is higher than the structural risk threshold, it is marked as a risk area.

[0013] Preferably, the node energy status includes the initial power of the node, the remaining power of the node, and the power supply status of the node's solar panel.

[0014] Preferably, a redundancy regularization term is introduced into the activation scoring function to indicate whether there are other nodes that have recently performed oversampling in the local area of ​​the node, so as to avoid repeated sampling.

[0015] Preferably, the activation threshold is a dynamic threshold, which is adjusted based on the average environmental fluctuation value of all nodes in the most recent time window in the past. When the average environmental fluctuation value is the largest, the activation threshold is the smallest.

[0016] Preferably, the acoustic wave reflection data includes an acoustic wave reflection signal, an acoustic wave main reflection delay, a main reflection echo intensity, an attenuation intensity, a number of multiple reflections and a count of effective reflection layers.

[0017] Preferably, a risk-sensitive regularization term is introduced into the structural mapping model to highlight the potential risk sensitivity of high stratification and near phase change temperature. It is constructed by the product of stratification complexity, snow temperature sensitivity and structural risk score. The stratification complexity is obtained by normalizing the number of multi-echo structures, and the snow temperature sensitivity is obtained by the average temperature value of the past time window.

[0018] Preferably, the scheduling threshold is a dynamic threshold, and the threshold is adjusted based on the temperature drop in the most recent time window. When the temperature drop is large, the scheduling threshold is small.

[0019] Preferably, when the node scheduling score is greater than the scheduling threshold, the node enters an active state and performs acoustic wave sampling; when the node scheduling score is less than or equal to the scheduling threshold, the node enters a sleep mode and delays the next sampling.

[0020] Preferably, the information aggregation module is further used to calculate the gradient change rate of the region to assist in determining whether there is a trend of structural drastic change, which is obtained by central difference calculation.

[0021] In summary, the technical effects and advantages of the present invention are as follows: The snow state monitoring system based on wireless sensing technology and cloud services has designed a monitoring mechanism that can perceive the multi-level structural state of the snow, so that the system no longer relies on indirect parameters such as single-point snow depth, but can identify key snow layer characteristics including layer thickness, compaction degree, and structural heterogeneity, thereby improving the ability to identify potential structural instability. Secondly, the system introduces an energy state perception model in the design of sensor nodes, and dynamically adjusts the node sampling frequency, data return behavior and communication path in combination with the changing trend of snow conditions, effectively extending the equipment operation cycle and improving maintainability in harsh environments. In addition, the system proposed in the present invention has a "risk-driven scheduling capability" that can prioritize monitoring resources based on the identified changes in the snow layer state, ensuring real-time upload and priority processing of data in high-risk areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a structural schematic diagram of the present invention. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0024] like Figure 1 As shown, a snow status monitoring system based on wireless sensing technology and cloud services includes:

[0025] A node deployment module, wherein the node deployment unit is used to record basic node information for each node and calculate a risk sensitivity score. The basic node information includes node location information, node terrain conditions, node energy status, and node slope conditions. The risk sensitivity score is obtained by weighted calculation of the basic node information;

[0026] A node activation module is used to determine whether to activate a node and evaluate the node using a node activation score calculated by weighting the node's basic information and risk sensitivity score. Nodes that reach the activation threshold are activated, and after activation, acoustic wave sampling and information recording are performed to obtain acoustic wave reflection data, node temperature, and snow depth.

[0027] A structure recognition module is used to model and identify characteristic structure vectors and output a structural risk score. The characteristic structure vectors include acoustic reflection data, node basic information, node temperature, and snow depth. The structural risk score is calculated using a structure mapping model composed of two layers of linear transformation and activation function.

[0028] A node scheduling module is used to determine whether each node continues sampling or enters sleep mode during the current scheduling cycle. The determination is made by comparing the node scheduling score with the scheduling threshold. The node scheduling score is obtained by weighted calculation of the structural risk score, node energy status, node location information, and temperature change rate input into the scheduling score function. The temperature change rate is calculated based on the node temperature.

[0029] An information aggregation module is used to aggregate the input data of all activated nodes to obtain a regional structural risk map. The input data includes the spatial coordinates of the nodes, the structural risk score, and the node scheduling status. The regional structural risk map is calculated using a spatial weighted interpolation algorithm and then written into a two-dimensional grid or vector map. A structural risk threshold is set. When the structural risk score is higher than the structural risk threshold, it is marked as a risk area.

[0030] The following is a detailed description of the system operation steps in this embodiment:

[0031] Step 1: Node modeling and risk perception presetting

[0032] This step is to complete the initialization of the entire system deployment and establish reasonable data support for the structure identification and dynamic scheduling in the subsequent steps. In particular, in this patent, the snow scene has the following typical characteristics:

[0033] Node deployment environments are diverse (such as mountains, rooftops, and bridges);

[0034] The power source relies on batteries and possibly solar energy, and long-term operation is extremely sensitive to energy consumption;

[0035] The goal is to prioritize the identification and monitoring of areas with high structural risk and high terrain sensitivity.

[0036] Record basic information for each node i and build a node status table

[0037]

[0038] x i ,y i : The latitude and longitude of the node, obtained by the GPS module;

[0039] h i : The node's altitude, which can be extracted through GPS or digital elevation map (DEM);

[0040] T i Node deployment terrain type, a discrete variable (e.g., roof = 1, slope = 2, flat ground = 3, mountain = 4), obtained through manual annotation or remote sensing image recognition;

[0041] The initial power of the node, in Wh, is read in real time by the battery power acquisition module;

[0042] S i : Whether there is solar panel power supply, Boolean variable (1 for yes, 0 for no);

[0043] V i : The slope angle at the node deployment location, in degrees (°), measured by the inclination sensor.

[0044] Calculate the risk sensitivity score λ for each node i , used to guide subsequent structural monitoring priorities:

[0045]

[0046] λ i : The risk sensitivity score of the node, normalized to [0,1]. The higher the value, the more likely it is to experience structural instability.

[0047] w1, w2, w3: artificially set weighting coefficients used to control the influence weights of different factors (for example, w1 = 0.6, w2 = 0.3, w3 = 0.1);

[0048] T i : Terrain type factor (discrete value);

[0049] V i : slope angle, unit is °;

[0050] SlopeFactor(h i ): A correction factor for slope and altitude, used to reflect that high altitude areas are more likely to have structural anomalies, such as:

[0051]

[0052] After the node deployment is completed, its status and risk sensitivity are uploaded to the cloud to form a node configuration data set

[0053]

[0054] Step 2: Structure-driven acoustic wave sampling scheduling

[0055] The goal of this step is to dynamically activate sensor nodes deployed in areas with higher snow disaster risk to perform structure-level acoustic wave sampling operations while meeting low power consumption constraints. This step is a key scheduling link in realizing the "state-level snow monitoring system" of this patent, and it follows the node risk sensitivity λ in step 1. i and energy status Using parameters such as the snow cover and the snowfall, an innovative scoring function and redundant regularization mechanism are used to implement a sampling decision-making strategy characterized by "structural risk driven + energy feasibility assessment + spatial coverage optimization." Unlike traditional timed or uniform strategies, snow cover states are highly terrain-dependent and subject to localized mutations. For example, snow on steep slopes is more prone to structural stratification and cavities, which can trigger avalanches; while changes in snow pressure at the base of power towers may indicate snow load risks. Therefore, this step not only considers whether the sensor has power but also answers the question of whether it is worth activating this node to sample structural information.

[0056] To achieve this goal, we define a node sampling activation scoring function A i (t), which is the core decision-making process of whether node t enters the acoustic wave sampling state at time t:

[0057]

[0058] in:

[0059] A i(t): activation score of node i at time t;

[0060] λ i : The risk sensitivity score of the node (generated by step 1), which reflects the physical importance of the node in structural monitoring;

[0061] E i (t): The current remaining power of the node, in Wh, which is periodically updated by the hardware power collection module;

[0062] The node's initial power has been set in step 1;

[0063] S i : Whether it has solar panel support (1 for yes, 0 for no);

[0064] Redundancy(x i ,y i ,R loc ): Indicates that the node is in the local area R loc (For example, a radius of 10 meters) whether there are other nodes that have recently performed oversampling. The function returns a density penalty term (for example, the number of nearest neighbor nodes divided by the total number of neighbors);

[0065] Volatility i (t): The node’s location in the nearest T v Estimates of the magnitude of minute-to-minute fluctuations in air temperature or snow surface height (e.g., temperature drop rate + snow depth fluctuations) to capture potentially drastic changes;

[0066] α, β, γ, δ, η: These are manually set adjustable weights. The default recommended values ​​are α = 0.4, β = 0.25, γ = 0.15, δ = 0.1, and η = 0.1.

[0067] The creativity of this formula is reflected in the following points:

[0068] Structural risk drives λ i : Unlike common scheduling strategies based on sensor power, this patent incorporates structural sensitivity into the core of sampling decision-making, emphasizing "who is worth activating rather than random activation";

[0069] Redundancy regularization term: evaluates the sampling density near the node through local inter-node status broadcast or cloud scheduling records to avoid repeated sampling and reduce the overall energy consumption of the system;

[0070] Volatility i (t): Used to identify areas with rapidly changing snow conditions (such as blizzard fronts and snowmelt areas under strong radiation), with strong scene adaptability and dynamic response capabilities;

[0071] Formula combinations are regular expressions rather than black box networks: suitable for embedded deployment, avoiding high computing power requirements and meeting the edge computing power limitations of nodes in snowy areas.

[0072] The system will automatically set an activation threshold τ(t) in each scheduling cycle. i Nodes with (t)>τ(t) are activated to perform sound wave sampling:

[0073]

[0074] τ(t): activation threshold of the system at time t;

[0075] τ0: system initial activation threshold (e.g. set to 0.5);

[0076] κ: adjustment coefficient;

[0077] Past T v Average Volatility of all nodes in the time window i (t) value, reflecting the degree of fluctuation of the current global environment.

[0078] When the system detects a drastic change in the overall environment (such as a sudden drop in temperature or concentrated snowmelt), τ(t) automatically decreases, allowing more nodes to enter the sampling state and implement a rapid response mechanism.

[0079] All meet A i The nodes with (t)>τ(t) are activated and perform the following specific operations:

[0080] Start the sonicator and emit short pulses in the frequency range of 1–3 kHz;

[0081] The microphone array collects the snow layer reflection signal S i (t), and record the echo characteristics (such as the main reflection delay Δt i , attenuation intensity A i , multiple reflection times n i );

[0082] Synchronously record local temperature, humidity, timestamp, etc. for subsequent model interpretation;

[0083] All results are input into the structural state identification model in the next step.

[0084] Step 3: Snow layer structure identification and multi-level risk scoring

[0085] This step inherits the acoustic reflection data S collected by the activated node i in the previous step i (t), and its associated structural feature Δt i (main reflection delay time), Ai (reflection intensity), n i (number of multiple reflections), T i (t)(current temperature), H i (t)(snow depth), etc., constitute the input feature vector x i By deeply modeling and identifying these structural signals, it is inferred whether there are structural anomalies (such as weak layers, cavities, ice plate interlayers, etc.) inside the snow body in the area, and thus a continuous structural risk score R is output. i , which is then used by subsequent behavioral scheduling and regional warning modules. Unlike traditional snow monitoring systems, which rely solely on representative data such as snow depth or temperature, this step involves constructing a "structural state mapping model" to exploit the hierarchical variation characteristics of acoustic wave propagation signals and substantively determine whether the snow has potentially unstable structures.

[0086] The input feature vector is defined as follows:

[0087] x i =(Δt i ,A i ,n i ,T i (t),H i (t))

[0088] Δt i : The main reflection delay of the acoustic wave, in ms, is obtained by detecting the acoustic wave timing signal and reflects the total thickness of the snow layer or the first reflection interface;

[0089] A i : The main reflection echo intensity, in dB, represents the density of the medium. Ice layers usually have strong reflections.

[0090] n i : Effective reflection layer count, identified in the echo signal by zero crossing / local peak method;

[0091] T i (t): the current temperature measured by the node, in °C;

[0092] H i (t): Current snow depth, obtained by the node's independent snow depth sensor, in cm.

[0093] This step uses a multi-layer acoustic structure fusion model Its design goal is to extract the implicit physical structure information of multi-layer snow bodies from limited sensor data and output a continuous risk score R i To adapt to the edge computing environment, the model uses a lightweight two-layer neural network + structural physics regularization. The overall expression is as follows:

[0094] R i=Sigmoid(W2·ReLU(W1·x i +b1)+b2)

[0095] R i : Structural risk score, defined as a continuous variable in the interval [0, 1], where higher values ​​indicate more unstable structures;

[0096] It is a structural mapping model, consisting of two layers of linear transformation and activation function, with a parameter set of θ = {W1, b1, W2, b2};

[0097] The model is trained in the cloud using historical structured label data and deployed on the edge of each node after model compression.

[0098] Compared with traditional classifiers, the output of this model is a continuous risk measure, which is more suitable for building a hierarchical response mechanism.

[0099] In order to enhance the model's ability to distinguish "dangerous structural features", especially when the number of echoes n i Multiple, reflection intensity A i When the fluctuation is large or the temperature is close to the critical melting point (for example, -1°C to 0°C), this step designs a snow temperature-driven risk-sensitive regularization term as an enhancement mechanism in the model training phase:

[0100]

[0101] BCE(R i ,y i ): basic cross entropy loss function, y i ∈{0,1} indicates whether the region actually experiences structural instability at that moment (experimental or historical data annotation);

[0102] The second term is the modulation term of number of structural layers × snow temperature sensitivity × risk score;

[0103] Indicates the past T w The minute average temperature, in °C, is the sensitivity of the control risk hot zone identification;

[0104] n i / 3: Normalizes the number of multi-echo structures, representing the complexity of snow layering. The larger the number, the higher the risk.

[0105] λ: regularization weight coefficient, recommended to be set to 0.15;

[0106] This regularization term highlights the potential risk sensitivity of the "high stratification + near phase transition temperature" situation, and is highly physically reasonable and scenario-appropriate.

[0107] Step 4: Node behavior scheduling driven by structure and energy synergy

[0108] The goal of this step is to score the structural risk R based on the output of the previous step 3. i and the node’s current remaining power E i (t), performs node behavior scheduling, and decides whether each node should continue sampling or enter a dormant state in the current scheduling cycle, so as to maximize the overall system operation time and energy efficiency while ensuring the structural risk response capability.

[0109] The input variables for this step include:

[0110] R i : The structural risk score from step 3, ranging from [0,1], represents the degree of structural instability of the area where node i is located;

[0111] E i (t): The remaining power of the current node, in Wh, read in real time by the node power collection module;

[0112] The initial power of the node;

[0113] x i ,y i : The geographical location of the node, used for redundant calculations;

[0114] The ambient temperature change rate ΔT(t) is obtained by the temperature sensor and is expressed in °C.

[0115] Node behavior scheduling is determined by a comprehensive scoring function U i (t) Control:

[0116]

[0117] Variable Description:

[0118] U i (t): The scheduling score of node i at time t, used to determine whether to activate;

[0119] α: structural risk weight (α=0.5 is recommended);

[0120] β: power weight (β=0.3 is recommended);

[0121] γ: Redundancy penalty weight (γ=0.2 is recommended);

[0122] Redundancy(x i ,y i ): Calculate whether there are other sampled nodes in the neighborhood of the node (such as radius r = 10m). It can be expressed as the normalized density of activated nodes per unit area, ranging from [0,1].

[0123] The system sets a dynamic scheduling threshold τ(t) to determine the sampling threshold in the current cycle, which is defined as follows:

[0124] τ(t)=τ0·(1-∈·ΔT(t))

[0125] Variable Description:

[0126] τ(t): dynamic scheduling threshold;

[0127] τ0: Initial threshold value (e.g. 0.5);

[0128] ∈: temperature fluctuation adjustment coefficient (e.g. 0.05);

[0129] ΔT(t): represents the temperature drop in the last T minutes (e.g., 3°C, τ(t) = 0.5·(1-0.05·3) = 0.425). It comes from node or cloud meteorological data.

[0130] The scheduling strategy execution process is as follows:

[0131] The node calculates its U locally i (t);

[0132] If U i (t)>τ(t), the node enters the active state and performs sound wave sampling;

[0133] If U i (t)≤τ(t), the node enters sleep mode and delays the next sampling;

[0134] The node behavior state can be stored as A i (t+1)∈{activate, sleep}, which serves as the input for subsequent system evaluation and scheduling.

[0135] Step 5: Cloud aggregation, risk area identification, and remote coordinated response

[0136] The task of this step is to check the data (x i ,y i ,R i ,A i (t+1)) is aggregated to construct a structural risk map H(x, y) for the entire monitoring area, which can be used for subsequent visualization, platform analysis, and auxiliary early warning functions. This step does not involve behavioral control and is only responsible for spatial aggregation and information expression.

[0137] All input data must meet A i (t+1)=1, that is, only the nodes that have been activated and successfully sampled are aggregated:

[0138] x i ,y i: The spatial coordinates of the node are derived from the deployed static GPS information;

[0139] R i : Structural risk score, derived from the model in step 3

[0140] A i (t+1): Node behavior status, derived from the scheduling strategy in step 4. A value of 1 indicates valid sampling.

[0141] All data is uploaded to the cloud platform in real time and written to the database or cache in a structured format (such as JSON / Protobuf, etc.).

[0142] To construct a continuous regional structural risk map H(x,y), a spatially weighted interpolation algorithm based on a Gaussian kernel is used, which is defined as follows:

[0143]

[0144] illustrate:

[0145] H(x,y): structural risk score of grid point (x,y);

[0146] R i : Node structure score from step 3;

[0147] d i : Euclidean distance from node i to (x,y);

[0148] σ: spatial attenuation coefficient (e.g. σ = 20m);

[0149] The set of activated nodes whose distance (x, y) is less than r (e.g. r = 50m);

[0150] The interpolation calculation results are written into a two-dimensional grid or vector map for risk heat mapping.

[0151] The cloud system sets the structural risk threshold R crit (e.g. 0.7), used to determine whether the grid is a high-risk area, marked with red highlight:

[0152] If H(x,y)>R crit , the system will mark the point as high risk in the results;

[0153] Visual charts can be output by overlaying layers for remote viewing by managers.

[0154] To assist managers in identifying sudden changes in regional status, the system also calculates the regional gradient change rate Ψ(x,yi):

[0155]

[0156] illustrate:

[0157] Ψ(x,y): The risk change intensity of the grid (x,y), used to assist in determining whether there is a trend of structural drastic change;

[0158] The values ​​are calculated by central difference;

[0159] Non-control parameters, only used for information analysis and heat map auxiliary layer drawing.

[0160] The output of this step includes:

[0161] Spatial risk map H(x,y) (for heat map display and regional analysis);

[0162] Auxiliary layer Ψ(x,y) (visualize gradient changes);

[0163] High-risk point list ((x,y) position + H(x,y) value);

[0164] The data is uploaded to the cloud database for platform management and access.

[0165] Final implementation notes:

[0166] All interpolation and gradient calculations are done asynchronously by cloud microservices;

[0167] The recommended update period for the heat map is every 5 minutes (configurable);

[0168] Density adaptation in interpolation areas: use fine meshes where node density is high, and reduce resolution appropriately in low-density areas;

[0169] The output results can be displayed through the front-end web platform or mobile map module, and support API access and download.

[0170] The technical solutions in the above-mentioned embodiments of the present application have at least the following technical effects or advantages: This solution designs a monitoring mechanism capable of sensing the multi-layered structural state within the snowpack. This enables the system to no longer rely on indirect parameters such as single-point snow depth, but instead to identify key snow layer characteristics including layer thickness, compaction level, and structural heterogeneity, thereby improving the ability to identify potential structural instability. Secondly, the system incorporates an energy state perception model into the sensor node design. Based on snow condition trends, it dynamically adjusts node sampling frequency, data return behavior, and communication paths, effectively extending the equipment's operating cycle and improving maintainability in harsh environments. Furthermore, the proposed system possesses a "risk-driven scheduling capability" that prioritizes monitoring resources based on identified snow layer state changes, ensuring real-time upload and priority processing of data in high-risk areas. The entire system is built on a widely deployable wireless sensor platform, combined with a cloud-based dynamic analysis engine and intelligent scheduling algorithm, achieving a complete closed-loop "near-ground precision perception - remote dynamic management."

[0171] The working principle is as follows: by setting up a monitoring mechanism to perceive the multi-level structural status of the snow, key snow layer characteristics including layer thickness, compaction degree, and structural heterogeneity are identified, an energy state perception model is introduced, and combined with the snow condition change trend, the node sampling frequency, data return behavior and communication path are dynamically adjusted.

[0172] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A snow status monitoring system based on wireless sensing technology and cloud services, characterized in that: include: A node deployment module, wherein the node deployment unit is used to record basic node information for each node and calculate a risk sensitivity score. The basic node information includes node location information, node terrain conditions, node energy status, and node slope conditions. The risk sensitivity score is obtained by weighted calculation of the basic node information; A node activation module is used to determine whether to activate a node and evaluate the node using a node activation score, which is calculated by weighting the node's basic information and risk sensitivity score. Nodes that reach the activation threshold are activated. After the nodes are activated, acoustic wave sampling and information recording are performed to obtain acoustic wave reflection data, node temperature, and snow depth. A structure recognition module, which is used to model and identify characteristic structure vectors and output a structure risk score; The characteristic structure vector includes acoustic wave reflection data, node basic information, node temperature and snow depth. The structural risk score is calculated by a structural mapping model, which consists of two layers of linear transformation and activation function. A node scheduling module is used to determine whether each node continues sampling or enters sleep mode during the current scheduling cycle. The determination is made by comparing the node scheduling score with the scheduling threshold. The node scheduling score is obtained by weighted calculation of the structural risk score, node energy status, node location information, and temperature change rate input into the scheduling score function. The temperature change rate is calculated based on the node temperature. An information aggregation module is used to aggregate the input data of all activated nodes to obtain a regional structural risk map. The input data includes the spatial coordinates of the nodes, the structural risk score, and the node scheduling status. The regional structural risk map is calculated using a spatially weighted interpolation algorithm and then written into a two-dimensional grid or vector map. Set a structural risk threshold. When the structural risk score is higher than the structural risk threshold, the area is marked as a risk area.

2. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: The node energy status includes the initial power of the node, the remaining power of the node, and the power supply status of the solar panel of the node.

3. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: A redundancy regularization term is introduced into the activation scoring function to indicate whether there are other nodes that have recently performed oversampling in the local area of ​​the node, so as to avoid repeated sampling.

4. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: The activation threshold is a dynamic threshold, which is adjusted based on the average environmental fluctuation value of all nodes in the most recent time window in the past. When the average environmental fluctuation value is the largest, the activation threshold is the smallest.

5. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: The acoustic wave reflection data includes an acoustic wave reflection signal, an acoustic wave main reflection delay, a main reflection echo intensity, an attenuation intensity, a number of multiple reflections, and a count of effective reflection layers.

6. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: A risk-sensitive regularization term is introduced into the structural mapping model to highlight the potential risk sensitivity of high stratification and near phase change temperature. It is constructed by multiplying the stratification complexity, snow temperature sensitivity and structural risk score. The stratification complexity is obtained by normalizing the number of multi-echo structures, and the snow temperature sensitivity is obtained by the average temperature value of the past time window.

7. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: The scheduling threshold is a dynamic threshold, which is adjusted based on the temperature drop in the most recent time window. When the temperature drop is large, the scheduling threshold is small.

8. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: When the node scheduling score is greater than the scheduling threshold, the node enters an active state and performs acoustic wave sampling; when the node scheduling score is less than or equal to the scheduling threshold, the node enters a dormant mode and delays the next sampling.

9. The snow status monitoring system based on wireless sensing technology and cloud services according to claim 1 is characterized in that: The information aggregation module is also used to calculate the gradient change rate of the region to assist in determining whether there is a trend of structural drastic change, which is obtained through central difference calculation.

Images