Construction high-altitude falling accident intelligent monitoring and early warning method and system

Abstract

The invention relates to the technical field of construction falling monitoring, in particular to an intelligent monitoring and early warning method and system for construction high-altitude falling accidents. Time step-based behavior mode branches are introduced, and regional distribution probabilities of different behavior mode branches are divided; the behavior mode branch comprises a high-altitude operation mode and an edge movement mode; constructing a fusion control vector according to the scene risk vector and the individual deviation vector, calculating a weight vector, and activating a mask to calculate an output channel vector; a fusion feature column vector is established, the falling risk probability is predicted through a full connection layer, whether early warning is triggered or not is judged, if yes, a falling point is calculated according to a free falling body, the rope length is calculated in combination with a safety rope anchor point, whether the impact speed and impact force are calculated or not is judged, the elasticity and damping of a safety rope are introduced to calculate applied force, and a first-order ordinary differential equation set is established; setting displacement and impact speed conditions, and calculating a final falling point by combining the anchor point vector, the displacement and the displacement in the rope direction of the operator.

Description

Technical Field

[0001] The present invention relates to the technical field of construction fall monitoring, and in particular to an intelligent monitoring and early warning method and system for construction high-altitude fall accidents. Background Art

[0002] Traditional intelligent monitoring for construction height fall accidents relies on rigid area coding (e.g., dividing the work area into zones A / B / C) and using fixed thresholds to determine whether sensor data (e.g., position and acceleration) exceeds limits. However, high-altitude work activities are highly dynamic and ambiguous: the same area (e.g., Zone A) may contain both resting personnel and high-risk workers, making it impossible to distinguish them with a single area label; geolocation codes become ineffective when personnel move across boundaries; and time-series variations in safety rope force and attitude angles are not fully explored, resulting in a high rate of false alarms and a significant risk of missed reports.

[0003] In the accident prediction stage, existing technologies mostly use a simplified free-fall model, ignoring the mechanical buffering mechanism of the safety rope (such as elastic deformation and impact fracture threshold). They are unable to distinguish between scenarios such as effective braking of the safety rope, rope breakage, and buffering shock, resulting in large deviations in fall predictions and delays in rescue deployment for workers and construction areas. Summary of the Invention

[0004] In response to the above-mentioned shortcomings of the existing technology, the present invention provides an intelligent monitoring and early warning method and system for high-altitude fall accidents during construction, which can effectively solve the problem that the existing technology is difficult to integrate with actual construction scenarios, resulting in poor accuracy in monitoring and early warning of fall accidents during construction.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: The present invention provides an intelligent monitoring and early warning method for construction height-falling accidents, which at least includes: Collect observation vectors of workers in the construction area and establish a single-moment state vector; Fall risk prediction using time series depth model, including: Obtain a multi-channel signal matrix based on the single-time state vector and output a context vector; Introducing time-step-based behavior pattern branches to divide the regional distribution probabilities of different behavior pattern branches; Behavioral mode branches include high-altitude operation mode and edge movement mode; Construct a fusion control vector based on the scene risk vector and the individual bias vector, and calculate the weight vector and activation mask to calculate the output channel vector; Establish a fusion feature column vector, predict the probability of falling risk through the fully connected layer, and determine whether to trigger an early warning. If triggered: Calculate the falling point based on free fall, calculate the rope length based on the safety rope anchor point, and determine whether to calculate the impact velocity and impact force; The elasticity and damping of the safety rope are introduced to calculate the applied force, and a set of first-order ordinary differential equations is established; Set the displacement and impact velocity conditions, and calculate the final fall point by combining the anchor point vector, displacement, and the operator's rope direction displacement.

[0006] The method of outputting the context vector is: Set a sliding time window; Construct a timing matrix and obtain a multi-channel signal matrix through transposition operation; Extract a local sequence from the input multi-channel signal matrix as the context segment of the time step; Perform one-dimensional convolution layer processing and output the context vector through the ReLU activation function.

[0007] The method for determining regional distribution probability is: For each time step, construct the regional distribution probability of the behavior at that time step among all pattern branches; The context vector is input into the two-layer perceptron MLP and then mapped into a real vector to obtain the regional distribution probability of the time walking behavior belonging to different behavior pattern branches.

[0008] The scenario risk vector is constructed through temperature, humidity, and wind speed; the individual deviation vector is constructed through fatigue and concentration.

[0009] Furthermore, the output channel vector is calculated as follows: Get all channel dimensions, extracted channel features, activation masks, fusion weights of behavior pattern branches, and probability calculation output channel vectors of behavior pattern branches.

[0010] If there is a time step under which the warning is triggered, the triggering time is determined, the difference between the triggering time and the current time is judged, and the free fall model is introduced to calculate the falling point.

[0011] Furthermore, the method for calculating the impact velocity and impact force is: Set the operator's mass, obtain the operator's velocity vector, and calculate the impact velocity when the safety rope is fully pulled by combining the gravity acceleration and the difference; The impact force is calculated based on the impact speed, mass and cushioning time.

[0012] Furthermore, the method for determining the applied force is: Calculate the difference between the current rope length and the rope length when it is stationary and unstretched, calculate the speed of rope length change, and input the elastic constant and damping coefficient of the safety rope to calculate the applied force.

[0013] The intelligent monitoring and early warning system for construction height-falling accidents is applied to the intelligent monitoring and early warning method for construction height-falling accidents, including: The state vector establishment module collects the observation vectors of workers in the construction area and establishes a single-time state vector; The fall risk prediction module obtains a multi-channel signal matrix based on the single-time state vector and outputs a context vector; Introducing time-step-based behavior pattern branches to divide the regional distribution probabilities of different behavior pattern branches; Behavioral mode branches include high-altitude operation mode and edge movement mode; Construct a fusion control vector based on the scene risk vector and the individual bias vector, and calculate the weight vector and activation mask to calculate the output channel vector; Establish a fusion feature column vector and predict the probability of falling risk through a fully connected layer; The fall point prediction module determines whether to trigger an early warning based on the fall risk probability. If triggered: Calculate the falling point based on free fall, calculate the rope length based on the safety rope anchor point, and determine whether to calculate the impact velocity and impact force; The elasticity and damping of the safety rope are introduced to calculate the applied force, and a set of first-order ordinary differential equations is established; Set the displacement and impact velocity conditions, and calculate the final fall point by combining the anchor point vector, displacement, and the operator's rope direction displacement.

[0014] A computer-readable storage medium stores a computer program, which implements the steps of the method when executed by a processor.

[0015] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects: By calculating the distribution probability of workers belonging to different behavior patterns at each time point, the ambiguity of behavior is quantified; an independent feature extraction channel is designed for each type of behavior pattern, and differentiated one-dimensional convolution kernels are used to capture the key features of specific patterns; finally, weights and masks are dynamically generated by combining environmental parameters and individual states, automatically activating relevant branches and suppressing irrelevant noise, thereby achieving fine-grained perception and monitoring of construction behavior.

[0016] By calculating the distance between the falling point and the safety rope anchor point, if the rope length threshold is exceeded, the impact force judgment is triggered. The instantaneous tension on the safety rope is calculated according to the mass of the personnel, the buffering time and the impact speed. It is determined whether the safety rope should be modeled as a spring damping system to simulate the reciprocating oscillation process of the personnel due to the elasticity of the rope. The motion trajectory is solved in real time through Newton's law until the speed or acceleration is lower than the static threshold. Finally, the precise falling point after the buffering is completed is output, providing a key basis for the layout of the rescue cushion and the evacuation of personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0018] Figure 1 Schematic diagram of the overall method of the present invention. DETAILED DESCRIPTION

[0019] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] The present invention will be further described below with reference to the embodiments.

[0021] Example 1 (see Figure 1 ): An intelligent monitoring and early warning method for construction height-falling accidents, comprising at least: Collect sensor observation vectors of workers in the construction area, including: vertical acceleration (Reflects the acceleration state of the operator's body or equipment in the vertical direction), acceleration change rate (vertical acceleration at the current moment minus vertical acceleration at the previous sampling moment), and the coordinate position of the operator (the operator's position), the operator's velocity vector , attitude angle (Including three-dimensional attitude angle information: roll angle, pitch angle, yaw angle, used to determine whether the person is tipping over, rolling, unbalanced, etc.), attitude angle change rate , Safety rope stress (Current force on the safety rope), rate of change of force on the safety rope , Indicates area codes, such as operating platforms, scaffolding edges, etc., and rigidly divides the entire high-altitude work area into several discrete, fixed area blocks (such as area A, area B, area C, etc.).

[0022] Then, the multi-class data collected above are standardized and scaled, and then constructed into a single-time state vector :

[0023] Indicates the operating personnel, Indicates the collection time; In this way, by integrating multiple data, it is possible to subsequently determine whether the operator has a tendency to fall.

[0024] Then, risk prediction is performed through the time series depth model, and we have: Setting a sliding time window , such as (e.g. 2 seconds, 100 frames); The construction length is , the time series matrix with a single moment state vector dimension of D ,Considering the complexity of construction behavior, the behavior within the window is heterogeneous, and unified modeling may fail. The time series matrix is ​​transposed to obtain a multi-channel signal matrix ; In risk monitoring for workers working at heights, relying solely on single-point data (such as the current position or sensor value) may not accurately determine whether there is a fall risk. This is because falls often have dynamic characteristics. Therefore, to avoid short-term fluctuations that interfere with judgment: From the input multi-channel signal matrix China-Israel time step) as the center, intercept a length of The local sequence of , as the context fragment of this time step: ,in, , Represents the half-width of the sliding window, and the formula Represents the index, which is a slice symbol of the tensor / matrix index and is used to take all elements. Indicates from arrive elements, the whole Expressed as taking a multi-channel signal matrix All rows of , and arrive Elements of (column elements); Then, a one-dimensional convolutional layer is performed Process and output the context vector through the ReLU activation function : , extract the state changes in the context of this time step and capture behavioral trends and environmental changes.

[0025] Traditional methods divide work scenarios using rigid area coding, which fails to reflect behavioral diversity and boundary ambiguity (it does not consider changes in human behavior and only divides regions based on geographic location; it cannot handle ambiguous behavior in border areas or when moving across regions). To address this, we introduce time-step-based behavioral pattern branches to divide the regional distribution probabilities of different behavioral pattern branches, including: For each time step (Sliding time window ), construct the regional distribution probability of the behavior of this time step among all pattern branches , , Indicates that at time step Next, the operator belongs to The probability of the behavior pattern branch; Context vector After inputting the two-layer perceptron MLP, it is mapped into a real vector , and obtain the regional distribution probability of the walking behavior belonging to different behavior pattern branches .

[0026] Because the behavior patterns of workers working at height vary significantly in different scenarios (e.g., near the edge of a platform, at height), it is not appropriate to use a single model for unified feature extraction. Forced unification would obscure subtle but critical regional characteristics, reducing the accuracy of subsequent risk identification.

[0027] For each operating area behavior mode branch (such as edge movement mode, high-altitude operation mode, etc.), an independent feature extraction branch is set up, and each feature extraction branch uses a separate one-dimensional convolution kernel (Including the number of input channels, the number of output channels, and the kernel width. The corresponding values ​​for edge movement mode are: 3, 16, and 5; the corresponding values ​​for high-altitude operation mode are: 5, 32, and 7). The features under the behavior pattern branch are extracted (edge ​​movement mode: lateral drift of position coordinates, horizontal component of velocity vector, and periodic fluctuation of safety rope force, thereby identifying slippage or stepping when walking along the edge; high-altitude operation mode: vertical acceleration, pitch angle and pitch angle mutation amplitude, safety rope force change rate, safety rope tension peak, thereby detecting continuous rollover caused by high-altitude instability). In other words, different construction blocks use corresponding convolution kernels to capture regional differences:

[0028] in, , Indicates the behavioral pattern branches, totaling ,Since the current time point may have different behavior patterns, multiple ,feature extraction branches are used in parallel to perform multi-angle ,extraction, thereby maintaining pattern discrimination and ensuring that the sensitivity to ,high-risk action features is not averaged.

[0029] In actual high-altitude operations, not all behavioral pattern branches corresponding to different areas (such as different construction locations or risk levels) are applicable to every operator or every operation scenario. For example: Some operators may be in the peripheral area, but their behavior is more similar to that of the central area.

[0030] Therefore, statically using all behavior pattern branches wastes computing resources and may lead to misjudgments or missed detections. This means that not all behavior pattern branches are reliable for the current operator and scenario. Therefore, it is necessary to dynamically select which branches to include in the fusion and assign different weights. In order to achieve dynamic selection and personalized adaptation of behavioral pattern branches in different regions, a fusion control vector is constructed based on the scene risk vector (reflecting the global state of the current working environment, such as temperature, humidity, wind speed, etc.) and the individual deviation vector (reflecting the current operator's own characteristics, such as fatigue or concentration, both of which are in the range of 0-1. Both can also be obtained by obtaining the operator's physiological data and identifying it through a neural network model, or by mapping various physiological data to the same scale for weighted calculation). , adjust which behavior pattern branches should be activated and their weights, specifically: Using a two-layer MLP network Mapped into a weight vector for each region behavior pattern branch (controlling the proportion of each branch in the final fusion) and an activation mask, including: The first layer extracts the intermediate representation ;

[0031] in, 、 Represents the weight matrix and bias vector of layer 1; The second layer generates two outputs: weight vector and activation mask:

[0032] Then, the fusion weights of all regions Softmax normalization is performed to convert the scores of all regions into probabilities. In this way, at each time step, an operator can belong to multiple behavior pattern branches at the same time, and each behavior pattern branch corresponds to a probability weight. Therefore, at each time point, the behavior of a single operator can belong to multiple behavior pattern branches at the same time, and each branch has a corresponding probability weight. Then Gumbel-Softmax sampling is used to convert the continuous activation signal into a discrete binary decision to activate the mask candidate value. (Unprocessed by the activation function) is approximately binarized to obtain an activation mask that can produce 0 / 1 ,prevent irrelevant or unreliable branches from participating in judgment, and reduce noise and computational burden.

[0033] For the Layer, time step , the fusion formula is:

[0034] in, Indicates the Layer time step The output channel vector of Here it is used to get all channel dimensions (channel 1 to channel channels), is a slice symbol of a tensor / matrix, used to retrieve the representation of all channels at that position, Indicates the The behavioral pattern branch is in Layer time step The extracted channel features, Indicates the The activation mask of a behavior pattern branch is used to determine whether the branch is activated, which is 1 or 0. Indicates the The fusion weight of each behavior pattern branch.

[0035] Then you can take The fused feature column vector ; Establish a fully connected layer (trained weight matrix and bias vector, including Sigmoid activation function, used to output probability value, all common), used to map the fused feature column vector to risk probability, so as to obtain the first The falling risk probability predicted by the time step is the sliding time window above Repeat the steps of the fully connected layer to obtain the fall risk probability at all time steps and establish a risk probability sequence. If there is a time step If the probability of falling risk is greater than or equal to the threshold, the early warning mechanism is triggered (a voice warning is broadcast to the operator's terminal and an early warning is issued to the control center).

[0036] Furthermore, if there is a future time step If the early warning mechanism is triggered, the time of triggering the early warning mechanism is determined, and the difference between the time and the current time is determined. , and establish a fall risk accident point model to predict the fall accident point of the operator, then: Calculate the theoretical falling point based on the free fall model (no safety rope, broken safety rope or not worn):

[0037] in, Indicates the operator's predicted fall point in the future (three-dimensional vector), 、 are all three-dimensional vectors, Represents the acceleration due to gravity (a three-dimensional vector). Based on the physical model of free fall motion, combined with the current speed and position, it predicts the worker's fall accident location at a future moment.

[0038] Furthermore, the coordinates of the anchor point at the other end of the safety rope worn by the operator are obtained. (3D vector), usually installed on a solid, reliable structure (such as a building beam, platform edge, or dedicated hanging point), to ensure that the safety rope can effectively support the worker. Consider whether the worker will trigger the maximum effective rope length threshold during the fall: Calculate the safety rope length:

[0039] in, Indicates the rope length, describing the distance between the falling point and the anchor point coordinates; like If the rope length is less than or equal to the rope length threshold, it means that the safety rope is not fully stretched and the operator will not continue to fall. is the actual falling point; like If the value is greater than the rope length threshold, it means the safety rope is fully drawn and the next step is to proceed: Considering the impact of a free fall with the safety rope fully stretched, whether the safety rope will exceed its bearing capacity, there are: Calculate the tension in the rope to determine whether it will break: Assume that the quality of the operator is , the impact speed at the moment the rope is fully pulled is :

[0040] impact force The estimate is:

[0041] in, Indicates the buffering time, which describes the time from the operator touching the safety rope in free fall to the safety rope being fully stressed, usually set to 0.2-0.5s; If the impact force If the impact force is greater than the threshold, the safety rope is judged to be broken and the operator's accident falling point is .

[0042] If the impact force If the impact force is less than or equal to the impact force threshold, the safety rope is determined not to be broken and the next step is elastic buffer modeling: When the safety rope is fully extended, it does not stop instantly. Instead, due to the elasticity and damping of the rope, the operator will experience a buffering process similar to spring vibration - first being pulled back by the rope, then bouncing back, and gradually stabilizing. In this embodiment, the safety rope is regarded as a one-dimensional spring-damping system, and the applied force is:

[0043] in, Indicates the force exerted by the safety rope on the operator. represents the elastic constant (stiffness) of the safety rope, Indicates the current rope length and the rope length when it is still and not stretched The difference between represents the damping coefficient, Indicates the speed of rope length change .

[0044] Furthermore, the equation of motion of the operator's center of mass along the length of the rope satisfies Newton's second law:

[0045] in, Indicates the worker's acceleration in the direction of the rope.

[0046] Convert to a system of first-order ordinary differential equations:

[0047] in, Indicates the speed of the operator; Set the initial conditions,

[0048] The simulation integration process continues until any of the following conditions are met: If the acceleration of the person at the current position or the person's moving speed is less than the stillness threshold, the movement is still; When the current time point in the simulation reaches the maximum simulation time, in order to prevent the simulation from running indefinitely, a maximum simulation time limit is set. When this value is reached, the simulation process is forced to end, avoiding the problem of calculating the actual position of the operator from the beginning of the fall to the final rest after the entire buffering process.

[0049] Then, the final falling point of the operator is:

[0050] in, Indicates the final falling point of the operator after the simulated buffer process. The unit vector representing the direction from the anchor point to the falling point, Indicates the operator's displacement along the rope when the simulation ends.

[0051] This will enable monitoring and early warning of worker fall accidents in the future, so that the control center can implement corresponding safety risk control, including but not limited to evacuating construction workers from the fall point, quickly installing rescue buffer equipment at the fall point, etc., thereby improving the ability to ensure the personal safety of workers.

[0052] The intelligent monitoring and early warning system for construction height-falling accidents is applied to the intelligent monitoring and early warning method for construction height-falling accidents, including: The state vector establishment module collects the observation vectors of workers in the construction area and establishes a single-time state vector; The fall risk prediction module obtains a multi-channel signal matrix based on the single-time state vector and outputs a context vector; Introducing time-step-based behavior pattern branches to divide the regional distribution probabilities of different behavior pattern branches; Behavioral mode branches include high-altitude operation mode and edge movement mode; Construct a fusion control vector based on the scene risk vector and the individual bias vector, and calculate the weight vector and activation mask to calculate the output channel vector; Establish a fusion feature column vector and predict the probability of falling risk through a fully connected layer; The fall point prediction module determines whether to trigger an early warning based on the fall risk probability. If triggered: Calculate the falling point based on free fall, calculate the rope length based on the safety rope anchor point, and determine whether to calculate the impact velocity and impact force; The elasticity and damping of the safety rope are introduced to calculate the applied force, and a set of first-order ordinary differential equations is established; Set the displacement and impact velocity conditions, and calculate the final fall point by combining the anchor point vector, displacement, and the operator's rope direction displacement.

[0053] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method when executing the computer program.

[0054] A computer readable storage medium stores a computer program, and the computer program implements the steps of the method when executed by a processor.

[0055] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the protection scope of the technical solutions of the embodiments of the present application.

Claims

1. An intelligent monitoring and early warning method for construction height-falling accidents, characterized in that: include: Collect observation vectors of workers in the construction area and establish a single-moment state vector; Fall risk prediction using time series depth model, including: Obtain a multi-channel signal matrix based on the single-time state vector and output a context vector; Introducing time-step-based behavior pattern branches to divide the regional distribution probabilities of different behavior pattern branches; Behavioral mode branches include high-altitude operation mode and edge movement mode; Construct a fusion control vector based on the scene risk vector and the individual bias vector, and calculate the weight vector and activation mask to calculate the output channel vector; Establish a fusion feature column vector, predict the probability of falling risk through the fully connected layer, and determine whether to trigger an early warning. If triggered: Calculate the falling point based on free fall, calculate the rope length based on the safety rope anchor point, and determine whether to calculate the impact velocity and impact force; The elasticity and damping of the safety rope are introduced to calculate the applied force, and a set of first-order ordinary differential equations is established; Set the displacement and impact velocity conditions, and calculate the final fall point by combining the anchor point vector, displacement, and the operator's rope direction displacement.

2. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 1 is characterized in that: The method of outputting the context vector is: Set a sliding time window; Construct a timing matrix and obtain a multi-channel signal matrix through transposition operation; Extract a local sequence from the input multi-channel signal matrix as the context segment of the time step; Perform one-dimensional convolution layer processing and output the context vector through the ReLU activation function.

3. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 1 is characterized in that: The method for determining the regional distribution probability is: For each time step, construct the regional distribution probability of the behavior at that time step among all pattern branches; The context vector is input into the two-layer perceptron MLP and then mapped into a real vector to obtain the regional distribution probability of the time walking behavior belonging to different behavior pattern branches.

4. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 1 is characterized in that: The scenario risk vector is constructed by temperature, humidity, and wind speed; Individual deviation vectors are constructed through fatigue and concentration.

5. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 1 is characterized in that: The calculation output channel vector is specifically as follows: Get all channel dimensions, extracted channel features, activation masks, fusion weights of behavior pattern branches, and probability calculation output channel vectors of behavior pattern branches.

6. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 1 is characterized in that: If there is a time step under which the warning is triggered, the triggering time is determined, the difference between the triggering time and the current time is judged, and the free fall model is introduced to calculate the falling point.

7. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 6 is characterized in that: The method for calculating the impact velocity and impact force is: Set the operator's mass, obtain the operator's velocity vector, and calculate the impact velocity when the safety rope is fully pulled by combining the gravity acceleration and the difference; The impact force is calculated based on the impact speed, mass and cushioning time.

8. The intelligent monitoring and early warning method for construction height-falling accidents according to claim 1 is characterized in that: The method for determining the applied force is: Calculate the difference between the current rope length and the rope length when it is stationary and unstretched, calculate the speed of rope length change, and input the elastic constant and damping coefficient of the safety rope to calculate the applied force.

9. An intelligent monitoring and early warning system for construction height-falling accidents, applied to the intelligent monitoring and early warning method for construction height-falling accidents according to any one of claims 1 to 8, characterized in that: include: The state vector establishment module collects the observation vectors of workers in the construction area and establishes a single-time state vector; The fall risk prediction module obtains a multi-channel signal matrix based on the single-time state vector and outputs a context vector; Introducing time-step-based behavior pattern branches to divide the regional distribution probabilities of different behavior pattern branches; Behavioral mode branches include high-altitude operation mode and edge movement mode; Construct a fusion control vector based on the scene risk vector and the individual bias vector, and calculate the weight vector and activation mask to calculate the output channel vector; Establish a fusion feature column vector and predict the probability of falling risk through a fully connected layer; The fall point prediction module determines whether to trigger an early warning based on the fall risk probability. If triggered: Calculate the falling point based on free fall, calculate the rope length based on the safety rope anchor point, and determine whether to calculate the impact velocity and impact force; The elasticity and damping of the safety rope are introduced to calculate the applied force, and a set of first-order ordinary differential equations is established; Set the displacement and impact velocity conditions, and calculate the final fall point by combining the anchor point vector, displacement, and the operator's rope direction displacement.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images