Device and method for improving personal safety of construction site workers

Abstract

The invention discloses a device and method for improving the personal safety of workers on a construction site. The device comprises: an information collection unit, which is used to measure the man-machine distance between the worker and the construction machinery and obtain the position information of the worker, and save and upload the data to the analysis and processing unit; the analysis and processing unit, which is used to receive the data uploaded by the information collection unit, and judge whether the man-machine distance is within the safe area through a judgment model. If it is found that the worker is in the buffer area, an optimized Markov prediction model is established. When the prediction model predicts that the worker is about to enter the dangerous area at the next moment, an instruction is immediately sent to the early warning unit; if it is found that the worker is in the dangerous area, an instruction is directly sent to the early warning unit; the early warning unit is used to turn on a buzzer and an alarm light to remind the worker that he is in the dangerous area after receiving the instruction from the analysis and processing unit. The invention can obtain the position of the worker in time, avoid entering the dangerous area of ​​the equipment, and improve the personal safety of the worker.

Description

Technical Field

[0001] The present invention relates to a device and method for improving personal safety, and in particular to a device and method for improving the personal safety of workers on a construction site, and belongs to the technical field of automatic safety judgment. Background Art

[0002] In recent years, construction site safety accidents have occurred frequently. When large machines are working, sometimes in order to save time, the machines are overloaded, causing machine failures and dangers. Some large equipment is aging and not regularly maintained, causing heavy objects to fall during the operation of the machine and injure pedestrians. At present, the methods to improve the personal safety of workers are:

[0003] Use a safety light curtain device; the safety light curtain usually consists of two parallel beams perpendicular to each other. When the worker's body enters the light curtain, the light beam is blocked, thereby triggering the safety device or stopping the dangerous movement. The safety light curtain is usually used in conjunction with an emergency stop device to ensure that the machine or equipment is stopped immediately in the event of danger. This method can ensure the safety of workers, but stopping the machine suddenly can cause damage to the machine itself due to the inertia caused by emergency braking.

[0004] Enhance workers' safety awareness; before construction workers enter the construction site, they must be educated and assessed on construction safety and fire protection knowledge. Workers who fail the assessment are prohibited from entering the construction site to participate in the construction; they must wear safety helmets and chin straps when entering the construction site, and use personal labor protection equipment correctly; they must wear personal protective equipment correctly during construction work, and are not allowed to use fire privately, and are strictly prohibited from working after drinking; personal protective equipment must be checked regularly and replaced in time when damaged; construction workers must wear clothing that meets the regulations of the construction site, and are strictly prohibited from wearing high heels, slippers, etc. when entering the construction site. This method only enhances workers' safety awareness, but there are many potential dangers in the complex environment of the construction site, and it is difficult for workers to judge their safety range in time.

[0005] Regularly inspect and repair large equipment, equip large machines with intelligent safety monitoring devices; use digital monitoring equipment to assist human monitoring of tower crane status; monitor the torque, amplitude, height, rotation, travel track and other limit parameters of the tower crane in real time, and issue early warnings and alarms when the limit is reached, assisting tower crane drivers to work safely and reducing the accident rate on the construction site. This method can effectively reduce accidents caused by machines, but if workers accidentally enter the dangerous area, it cannot remind workers to stay away from the dangerous area.

[0006] Many large equipment on the construction site bring many potential dangers to the construction site. Workers sometimes step into the dangerous area without knowing it, leading to accidents. Therefore, on the construction site, a safe distance automatic identification device is needed to judge the distance between people and machines and remind construction personnel and workers in time. Summary of the invention

[0007] Purpose of the invention: The purpose of the present invention is to provide a device and method for automatically identifying and judging the distance between a person and a machine and promptly reminding the workers to improve the personal safety of workers on the construction site.

[0008] Technical solution: A device for improving the personal safety of workers on a construction site according to the present invention comprises:

[0009] An information collection unit, including an infrared distance sensor, is used to measure the distance between the worker and the construction machinery and obtain the worker's position information, and save and upload the data to the analysis and processing unit;

[0010] The analysis and processing unit includes a cloud data platform, and is used to receive the data uploaded by the information collection unit and perform analysis and processing. The analysis and processing includes: judging whether the distance between the human and the machine is within the safe area through a judgment model. If the worker is found to be in the buffer area, an optimized Markov prediction model is established to predict the direction of the worker at the next moment. When the prediction model predicts that the worker is about to enter the dangerous area at the next moment, an instruction is immediately sent to the early warning unit. If the worker is found to be in the dangerous area, an instruction is directly sent to the early warning unit;

[0011] The early warning unit is used to turn on the buzzer and the alarm light to remind the workers that they are in the danger zone after receiving the instruction from the analysis and processing unit, and turn off the buzzer and the alarm light after receiving the instruction from the workers to leave the danger zone.

[0012] Furthermore, the infrared ranging sensor includes an infrared transmitter and an infrared receiver;

[0013] The infrared transmitter is used to transmit an infrared signal of a predetermined frequency.

[0014] The infrared receiver is used to collect the signal reflected after colliding with the obstacle, and perform calculations, and save and upload the data to the analysis and processing unit.

[0015] Furthermore, the judging model is used to judge whether the human-machine distance is within the safe area, specifically:

[0016] Assume that the calculation formula of the man-machine distance d is as follows:

[0017]

[0018] Among them, r i ,y i and z i are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the worker, respectively. 机 ,y 机 and z 机 They are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the construction machinery;

[0019] Assume the worker safety distance is L, and set:

[0020] L=s+x

[0021] In the formula, s is the dangerous distance of the construction machinery, and x is a preset constant;

[0022] When d>L, the worker is judged to be in the safe area. When s<d≤L, the worker is judged to have entered the buffer zone and may be in danger. An optimized Markov prediction model is established to judge the direction of the worker at the next moment. When d≤s, the worker is judged to be in the dangerous area.

[0023] Furthermore, the optimized Markov prediction model is specifically implemented by:

[0024] Constructing Markov prediction models;

[0025] Improve the Markov model and optimize parameters.

[0026] Furthermore, the constructing of the Markov prediction model includes:

[0027] S1.1: Expand the Markov model to a five-element vector group:

[0028] λ={S,V,π,A,B}

[0029] Among them, S is the state space, V is the observation symbol space, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation value matrix;

[0030] S1.2: The data obtained by the infrared sensor is used to determine the position of the worker, so that the worker can be reminded in time; set the state space:

[0031] S={s 1 ,s 2 ,s 3}

[0032] s 1 Indicates that the worker is in a safe area and is not in danger at the moment; 2 Indicates that the worker is in the buffer zone and may step into the danger zone; 3 Indicates that the worker is within the danger zone of a large machine and needs to be alerted immediately;

[0033] S1.3: Set up the node state sequence: Q = {q 1 ,q 2 , ..., q r}, q 1 ,q 2 , ..., qr ∈S, number of states N = 3;

[0034] S1.4: Determine the danger level of the worker’s current position by the distance between the worker and the machine, and set up an observation symbol space: V = {v 1 , v 2 , v 3 , v 4 , v 5},v 1 , v 2 , v 3 , v 4 and v 5 Indicates the degree of danger of the worker's position, the number of observations M = 5, the larger the subscript value, the more dangerous the worker's position;

[0035] S1.5: Set up an observation sequence: O = {o 1 , o 2 , ..., o r},o 1 , o 2 ,... r ∈V; a single observation value represents the location information of the worker observed within a specified time period;

[0036] S1.6: Establish initial state probability distribution: π={π 1， π 2 , π 3}, π i Indicates that the worker's initial state is s i The probability π i =P(q 1 =s 1 ), P is the probability calculation symbol, ∑π i =1(1≤i≤N);

[0037] S1.7: Establish state transfer matrix: A = [a ij ] N×N , a ij Indicates that the worker is from state s i Transfer to j The probability of ij =P(q t+1 =s j |q t =s i ),∑a ij =1(1≤j≤N);

[0038] S1.8: Establish the observation matrix: b i (v p ) indicates that the worker position is s i The observed value is vp The probability b i (v p )=P(o t =v p |q t =s i ), and ∑b j (v p )=1(1≤p≤M);

[0039] S1.9: Based on the forward-backward algorithm, update the Markov model parameters to maximize P(O|λ), where Specific:

[0040] π * =f t (i)

[0041]

[0042] Where, 1≤i, j≤M, T is time, f t () is the forward probability at the tth moment of the forward-backward algorithm, ε t () The backward probability of the forward-backward algorithm at time t;

[0043]

[0044] Among them, 1≤i≤M, 1≤k≤N;

[0045] After updating the parameters using the above method, we get P(O|λ * )>P(O|λ), and perform multiple iterations to obtain the maximum likelihood estimate of the Markov model.

[0046] Furthermore, the Markov model is improved and parameters are optimized, specifically, the HMM-SA-BW algorithm is improved based on the Markov model based on the simulated annealing algorithm and the Baum Welch algorithm to maximize P(O|λ), and the method includes:

[0047] S2.1: Initialize model parameters: π, A, B, annealing cooling temperature T 0 , temperature cooling coefficient k, termination temperature T * ;

[0048] Set the convergence condition ψ=-10 -2 , set the cooling function: T m+1 =kT m , where 0<k<1, T m represents the temperature at time m;

[0049] S2.2: Generate M×N independent random variables x that satisfy the normal distribution. The expectation of the normal distribution is 0 and the variance is Let a ik * =a ik +x, 1≤i≤M, 1≤k≤N, when a ik * When a is less than 0, let a ik * =0, the unit is normalized to obtain

[0050] S2.3: Update the model and introduce the Momentum first-order optimization acceleration algorithm:

[0051]

[0052] Among them, r t is the momentum variable, γ is the momentum hyperparameter, 0≤γ≤1, η t is the learning rate at the tth step, θ is the parameter P(O|λ) that needs to be updated, is the gradient of the loss function with respect to the parameters;

[0053] S2.4: Determine whether P(O|λ) satisfies the termination temperature or convergence condition: When P(O|λ) satisfies the condition, P(O|λ) is the optimal solution and the algorithm terminates; otherwise, return to S2.1 to continue.

[0054] Based on the same inventive concept, the present invention also provides a method for improving the personal safety of workers on a construction site, comprising:

[0055] Through the information collection unit, the distance between the worker and the construction machinery is measured and the worker's location information is obtained, and the data is saved and uploaded to the analysis and processing unit;

[0056] The analysis and processing unit receives the data uploaded by the information collection unit and performs analysis and processing, wherein the analysis and processing includes: judging whether the distance between the human and the machine is within the safe area through a judgment model; if the worker is found to be in the buffer area, establishing an optimized Markov prediction model to predict the direction of the worker at the next moment; when the prediction model predicts that the worker is about to enter the dangerous area at the next moment, immediately sending an instruction to the early warning unit; if the worker is found to be in the dangerous area, directly sending an instruction to the early warning unit;

[0057] After receiving the instruction from the analysis and processing unit, the early warning unit turns on the buzzer and the alarm light to remind the workers that they are in the danger zone, and turns off the buzzer and the alarm light after receiving the instruction from the workers to leave the danger zone.

[0058] Furthermore, the information acquisition unit includes an infrared distance measuring sensor, and the infrared distance measuring sensor includes an infrared transmitter and an infrared receiver;

[0059] The infrared transmitter is used to transmit an infrared signal of a predetermined frequency.

[0060] The infrared receiver is used to collect the signal reflected after colliding with the obstacle, and perform calculations, and save and upload the data to the analysis and processing unit.

[0061] Furthermore, the judging model is used to judge whether the human-machine distance is within the safe area, specifically:

[0062] Assume that the calculation formula of the man-machine distance d is as follows:

[0063]

[0064] Among them, r i ,y i and z i are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the worker, respectively. 机 ,y 机 and z 机 They are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the construction machinery;

[0065] Assume the worker safety distance is L, and set:

[0066] L=s+x

[0067] In the formula, s is the dangerous distance of the construction machinery, and x is a preset constant;

[0068] When d>L, the worker is judged to be in the safe area. When s<d≤L, the worker is judged to have entered the buffer zone and may be in danger. An optimized Markov prediction model is established to judge the direction of the worker at the next moment. When d≤s, the worker is judged to be in the dangerous area.

[0069] Furthermore, the optimized Markov prediction model is specifically implemented by: constructing a Markov prediction model, improving the Markov model and optimizing parameters;

[0070] The constructing of the Markov prediction model comprises:

[0071] Expand the Markov model to a five-element vector group:

[0072] λ={S,V,π,A,B}

[0073] Among them, S is the state space, V is the observation symbol space, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation value matrix;

[0074] The data obtained by the infrared sensor can be used to determine the position of the worker, so that the worker can be reminded in time; set the state space:

[0075] S={s 1 ,s 2 ,s 3}

[0076] s 1 Indicates that the worker is in a safe area and is not in danger at the moment; 2 Indicates that the worker is in the buffer zone and may step into the danger zone; 3 Indicates that the worker is within the danger zone of a large machine and needs to be alerted immediately;

[0077] Set up the node state sequence as: Q = {q 1 ,q 2 , ..., q r}, q 1 ,q 2 , ..., q r ∈S, number of states N = 3;

[0078] The danger level of the worker’s current position is determined by the distance between the worker and the machine, and an observation symbol space is set up: V = {v 1 , v 2 , v 3 , v 4 , v 5},v 1 , v 2 , v 3 , v 4 and v 5 Indicates the degree of danger of the worker's position, the number of observations M = 5, the larger the subscript value, the more dangerous the worker's position;

[0079] Set up an observation sequence: O = {o 1 , o 2 , ....o r},o 1， o 2 ,... r ∈V; a single observation value represents the location information of the worker observed within a specified time period;

[0080] Establish the initial state probability distribution: π = {π 1 , π 2 , π 3}, π i Indicates that the worker's initial state is s i The probability π i =P(q 1 =s 1), P is the probability calculation symbol, ∑π i =1(1≤i≤N);

[0081] Establish the state transfer matrix: A = [a ij ] N×N , a ij Indicates that the worker is from state s i Transfer to j The probability of ij =P(q t+1 =s j |q t =s i ),∑a ij =1(1≤j≤N);

[0082] Establish the observation matrix: B = [b i (v p )] N×M , b i (v p ) indicates that the worker position is s i The observed value is v p The probability b i (v p )=P(o t =v p |q t =s i ), and ∑b j (v p )=1(1≤p≤M);

[0083] Based on the forward-backward algorithm, the Markov model parameters are updated to maximize P(O|λ), where Specific:

[0084] π * =f t (i)

[0085]

[0086] Where, 1≤i, j≤M, T is time, f t () is the forward probability at the tth moment of the forward-backward algorithm, ε t () The backward probability of the forward-backward algorithm at time t;

[0087]

[0088] Among them, 1≤i≤M, 1≤k≤N;

[0089] After updating the parameters using the above method, we get P(O|λ *)>P(O|λ), perform multiple iterations to obtain the maximum likelihood estimate of the Markov model;

[0090] The Markov model is improved and parameters are optimized, specifically, the HMM-SA-BW algorithm is improved based on the Markov model based on the simulated annealing algorithm and the Baum Welch algorithm to maximize P(O|λ), and the method includes:

[0091] Initialize model parameters: π, A, B, annealing cooling temperature T 0 , temperature cooling coefficient k, termination temperature T * ;

[0092] Set the convergence condition ψ=-10 -2 , set the cooling function: T m+1 =kT m , where 0<k<1, T m represents the temperature at time m;

[0093] Generate M×N independent random variables x that satisfy the normal distribution. The expectation of the normal distribution is 0 and the variance is Let a ik * =a ik +x, 1≤i≤M, 1≤k≤N, when a ik * When a is less than 0, let a ik * =0, the unit is normalized to obtain

[0094] Update the model and introduce the Momentum first-order optimization acceleration algorithm:

[0095]

[0096] Among them, r t is the momentum variable, γ is the momentum hyperparameter, 0≤γ≤1, η t is the learning rate at the tth step, θ is the parameter P(O|λ) that needs to be updated, is the gradient of the loss function with respect to the parameters;

[0097] Determine whether P(O|λ) satisfies the termination temperature or convergence condition: When P(O|λ) satisfies the condition, P(O|λ) is the optimal solution and the algorithm terminates; otherwise, it returns to the step of initializing the model parameters and continues.

[0098] Beneficial effects: Compared with the prior art, the present invention utilizes the worker's location information to automatically determine whether the distance between humans and machines is within the safe area through a judgment model, effectively avoiding the situation where workers are not clear about the safe area of ​​large equipment and mistakenly enter; the present invention utilizes the optimized Markov prediction model to predict the direction of the workers at the next moment, greatly improving the accuracy of predicting the next direction of the workers. When the workers step into the dangerous area, early warning is immediately provided to the construction personnel and workers, reducing the accident rate on the construction site; the present invention can quickly judge, predict and issue early warning signals in the first time, greatly improving the personal safety of workers and reducing losses on the construction site. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 is a schematic diagram of a device unit according to an embodiment of the present invention;

[0100] Figure 2 Schematic diagram of a device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0101] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0102] As attached Figure 1 As shown, the device for improving the personal safety of workers on a construction site of this embodiment includes:

[0103] An information collection unit, including an infrared distance sensor, is used to measure the distance between the worker and the construction machinery and obtain the worker's position information, and save and upload the data to the analysis and processing unit;

[0104] The analysis and processing unit includes a cloud data platform, and is used to receive the data uploaded by the information collection unit and perform analysis and processing. The analysis and processing includes: judging whether the distance between the human and the machine is within the safe area through a judgment model. If it is found that the worker is in the buffer area, that is, the worker may be in danger, then an optimized Markov prediction model is established to predict the direction of the worker at the next moment. When the prediction model predicts that the worker is about to enter the dangerous area at the next moment, an instruction is immediately sent to the early warning unit. If the worker is found to be in the dangerous area, an instruction is directly sent to the early warning unit;

[0105] The early warning unit is used to receive instructions from the analysis and processing unit. When a worker enters a dangerous area, the early warning unit turns on a buzzer and an alarm light to remind the worker that the worker is in the dangerous area, and turns off the buzzer and the alarm light after receiving instructions from the worker to leave the dangerous area.

[0106] Furthermore, the infrared ranging sensor is mainly composed of an infrared transmitter and an infrared receiver. The infrared transmitter emits an infrared signal of a certain frequency. When the signal hits a large machine and is reflected, the receiver collects the reflected signal and performs calculations, while saving and uploading the data to the cloud data platform.

[0107] Furthermore, a judgment model is used to calculate the distance between man and machine through the distance formula between man and machine and combine it with the safe area range of large machines to determine whether the worker is in the danger zone. The distance formula between man and machine is as follows:

[0108]

[0109] Where d is the distance between the human and the machine, r i ,y i , z i are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the worker, respectively. 机 ,y 机 , z 机 They are respectively the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the large machine.

[0110] Table 1 shows the safety distances of some large machines:

[0111] Table 1 Safety distance of large machines

[0112] Common equipment Safety distance s / m excavator 5～10 Tower Crane 1.5 times the cantilever Concrete mixer 10 crane 5 Forklift 2.5

[0113] The formula for the human-machine safety distance is: L=s+2. When and only when d>L, the worker can be considered to be in a safe area. When the human-machine distance s<d≤L, it is considered that the worker has entered the buffer zone and may be in danger. An optimized Markov prediction model is immediately established to judge the direction of the worker at the next moment. When d≤s, the worker is determined to be in a dangerous area.

[0114] Furthermore, the Markov prediction model is used to predict the next move of the workers who enter the buffer zone, including the following steps:

[0115] 1.1 Due to the complexity of the construction site environment, if the worker's position is evaluated by model analysis, the evaluation becomes more complicated. Therefore, a Markov model is established for the worker's direction. The distance between man and machine and the position are used as observation symbols in the model to measure the worker's position and then determine the worker's direction. The Markov model is now expanded into a five-element vector group:

[0116] λ={S,V,π,A,B}.

[0117] 1.2 The data obtained by the infrared sensor is used to determine the position of the worker, so that the worker can be reminded in time. Set the state space: S = {s 1 ,s 2 ,s 3}

[0118] s 1 Indicates that the worker is in a safe area, that is, the worker is not in danger at the moment;2 Indicates that the worker is in the buffer zone, that is, the worker may step into the dangerous area at the next moment; 3 Indicates that the worker is within the danger range of a large machine and needs to be alerted immediately.

[0119] 1.3 Set up the node state sequence: Q = {q 1 ,q 2 , ..., q r}, q 1 ,q 2 , ..., q r ∈S, number of states N = 3.

[0120] 1.4 Determine the danger level of the worker’s current position by the distance between the worker and the machine, and set up an observation symbol space: V = {v 1 , v 2 , v 3 , v 4 , v 5},v 1 , v 2 , v 3 , v 4 , v 5 It indicates the danger level of the worker’s position. The observation number M=5. The larger the subscript value, the more dangerous the worker’s position is.

[0121] 1.5 Set up the observation sequence: O = {o 1 , o 2 ,... r},o 1 , o 2 ,…o r ∈V. An observation value represents the location information of the worker observed within a specified time period.

[0122] 1.6 Establish the initial state probability distribution: π = {π 1 , π 2 , π 3}, π i Indicates that the worker's initial state is s i The probability π i =P(q 1 =s 1 ),∑π i =1(1≤i≤N).

[0123] 1.7 Establish the state transfer matrix: A = [a ij ] N×N , a ij Indicates that the worker is from state s i Transfer to j The probability of ij =P(qt+1 =s j |q t =s i ),∑a ij =1(1≤j≤N)

[0124] 1.8 Establish the observation matrix: b i (v p ) indicates that the worker position is s i The observed value is v p The probability b i (v p )=P(o t =v p |q t =s i ), and ∑b i (v p )=1(1≤p≤M)

[0125] 1.9 Based on the forward-backward algorithm, update the Markov model parameters so that P(O|λ) reaches the maximum value, where The specific steps are as follows:

[0126] π * =f t (i)

[0127]

[0128] Where 1≤i, j≤M

[0129]

[0130] Among them, 1≤i≤M, 1≤k≤N

[0131] After updating the parameters using the above method, we get P(O|λ * )>P(O|λ), a maximum likelihood estimate of the Markov model can be obtained after multiple iterations, but the training method may fall into a local optimal state.

[0132] In order to avoid falling into the local optimal state, it is necessary to improve the Markov model and optimize the parameters: Based on the Markov model, the simulated annealing algorithm and the Baum Welch algorithm are improved to obtain the HMM-SA-BW algorithm to maximize P(O|λ). The steps are as follows:

[0133] 2.1 Initialize model parameters: π, A, B, annealing cooling temperature T 0 , temperature cooling coefficient k, termination temperature T * .

[0134] Set the convergence condition ψ=-10-2 , set the cooling function: T m+1 =kT m , where 0<k<1, T m Represents the temperature at time m.

[0135] 2.2 Generate M×N independent random variables x that satisfy the normal distribution. The expectation of the normal distribution is 0 and the variance is Let a ik * =a ik +x, 1≤i≤M, 1≤k≤N, when a ik * When a is less than 0, let a ik * =0, the unit is normalized to obtain

[0136] 2.3 To ensure that the model can quickly predict the next move of the workers and improve the safety of the workers, it is necessary to speed up the convergence of the algorithm. Therefore, the model is updated and the Momentum first-order optimization acceleration algorithm is introduced:

[0137]

[0138] Among them, r t is the momentum variable, γ is the momentum hyperparameter, 0≤γ≤1, η t is the learning rate at the tth step, θ is the parameter P(O|λ) that needs to be updated, is the gradient of the loss function with respect to the parameters.

[0139] 2.4 Determine whether P(O|λ) satisfies the termination temperature or convergence condition: When P(O|λ) satisfies the condition, P(O|λ) is the optimal solution and the algorithm terminates; otherwise, return to 2.1 and continue.

[0140] This algorithm will inevitably reach a point in a finite time where the solution does not change significantly within multiple steps, which means that the algorithm process is asymptotically convergent, that is, there is a global approximate optimal solution, which is the predicted next step for the worker.

[0141] Based on the same inventive concept, this embodiment also provides a method for improving the personal safety of workers on a construction site, comprising:

[0142] Through the information collection unit, the distance between the worker and the construction machinery is measured and the worker's location information is obtained, and the data is saved and uploaded to the analysis and processing unit;

[0143] The analysis and processing unit receives the data uploaded by the information collection unit and performs analysis and processing, wherein the analysis and processing includes: judging whether the distance between the human and the machine is within the safe area through a judgment model; if it is found that the worker is in the buffer area, an optimized Markov prediction model is established to predict the direction of the worker at the next moment; when the prediction model predicts that the worker is about to enter the dangerous area at the next moment, an instruction is immediately sent to the early warning unit; if it is found that the worker is in the dangerous area, an instruction is directly sent to the early warning unit;

[0144] After receiving the instruction from the analysis and processing unit, the early warning unit turns on the buzzer and the alarm light to remind the workers that they are in the danger zone, and turns off the buzzer and the alarm light after receiving the instruction from the workers to leave the danger zone.

[0145] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A device to improve the personal safety of workers on construction sites, It is characterized in that include: An information collection unit, including an infrared distance sensor, is used to measure the distance between the worker and the construction machinery and obtain the worker's position information, and save and upload the data to the analysis and processing unit; The analysis and processing unit includes a cloud data platform, which is used to receive the data uploaded by the information collection unit and perform analysis and processing. The analysis and processing includes: judging whether the distance between the human and the machine is within the safe area through a judgment model. If the worker is found to be in the buffer area, an optimized Markov prediction model is established to predict the direction of the worker at the next moment. When the prediction model predicts that the worker is about to enter the dangerous area at the next moment, an instruction is immediately sent to the early warning unit; if the worker is found to be in the dangerous area, an instruction is directly sent to the early warning unit; The early warning unit is used to turn on the buzzer and the alarm light to remind the workers that they are in the dangerous area after receiving the instruction from the analysis and processing unit, and turn off the buzzer and the alarm light after receiving no more instructions; The optimized Markov prediction model is specifically implemented as follows: Constructing Markov prediction models; Improve the Markov model and optimize parameters; The constructing of the Markov prediction model comprises: S1.1: Expand the Markov model to a five-element vector group: λ={S,V,π,A,B} Among them, S is the state space, V is the observation symbol space, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation value matrix; S1.2: The data obtained by the infrared sensor is used to determine the position of the worker, so that the worker can be reminded in time; set the state space: S={s 1 ,s 2 ,s 3 } s 1 Indicates that the worker is in a safe area and is not in danger at the moment; 2 Indicates that the worker is in the buffer zone and may step into the danger zone; 3 Indicates that the worker is within the danger zone of a large machine and needs to be alerted immediately; S1.3: Set up the node state sequence: Q = {q 1 ,q 2 , ..., q r }, q 1 ,q 2 , ..., q r ∈S, number of states N = 3; S1.4: Determine the danger level of the worker’s current position by the distance between the worker and the machine, and set up an observation symbol space: V = {v 1 , v 2 , v 3 , v 4 , v 5 },v 1 , v 2 , v 3 , v 4 and v 5 Indicates the degree of danger of the worker's position, the number of observations M = 5, the larger the subscript value, the more dangerous the worker's position; S1.5: Set up an observation sequence: O = {o 1 , o 2 ,... r },o 1 , o 2 ,... r ∈V; a single observation value represents the location information of the worker observed within a specified time period; S1.6: Establish initial state probability distribution: π={π 1 , π 2 , π 3 }, π i Indicates that the worker's initial state is s i The probability π i =P(q 1 =s 1 ), P is the probability calculation symbol, ∑π i =1(1≤i≤N); S1.7: Establish state transfer matrix: A = [a ij ] N×N , a ij Indicates that the worker is from state s i Transfer to j The probability of ij =P(q t+1 =s j |q t =s i ),∑a ij =1(1≤j≤N); S1.8: Establish observation matrix: B = [b i (v p )] N×M , b i (v p ) indicates that the worker position is s i The observed value is v p The probability b i (v p )=P(o t =v p |q t =s i ), and ∑b i (v p )=1(1≤p≤M); S1.9: Based on the forward-backward algorithm, update the Markov model parameters to maximize P(O|λ), where Specific: p * =f t (i) Where, 1≤i, j≤M, T is time, f t () is the forward probability at the tth moment of the forward-backward algorithm, ε t () The backward probability of the forward-backward algorithm at time t; Among them, 1≤i≤M, 1≤k≤N; After updating the parameters using the above method, we get P(O|λ * )>P(O|λ), and perform multiple iterations to obtain the maximum likelihood estimate of the Markov model; The Markov model is improved and parameters are optimized, specifically, the HMM-SA-BW algorithm is improved based on the Markov model based on the simulated annealing algorithm and the Baum Welch algorithm to maximize P(O|λ), and the method includes: S2.1: Initialize model parameters: π, A, B, annealing cooling temperature T 0 , temperature cooling coefficient k, termination temperature T * ; Set the convergence condition ψ=-10 -2 , set the cooling function: T m+1 =kT m , where 0<k<1, T m represents the temperature at time m; S2.2: Generate M×N independent random variables x that satisfy the normal distribution. The expectation of the normal distribution is 0 and the variance is Let a ik * =a ik +x, 1≤i≤M, 1≤k≤N, when a ik * When a is less than 0, let a ik * =0, the unit is normalized to obtain S2.3: Update the model and introduce the Momentum first-order optimization acceleration algorithm: Among them, r t is the momentum variable, γ is the momentum hyperparameter, 0≤γ≤1, η t is the learning rate at the tth step, θ is the parameter P(O|λ) that needs to be updated, is the gradient of the loss function with respect to the parameters; S2.4: Determine whether P(O|λ) satisfies the termination temperature or convergence condition: When P(O|λ) satisfies the condition, P(O|λ) is the optimal solution and the algorithm terminates; otherwise, return to S2.1 to continue.

2. The device for improving the personal safety of workers on a construction site according to claim 1, It is characterized in that The infrared distance measuring sensor comprises an infrared transmitter and an infrared receiver; The infrared transmitter is used to transmit an infrared signal of a predetermined frequency. The infrared receiver is used to collect the signal reflected after colliding with the obstacle, and perform calculations, and save and upload the data to the analysis and processing unit.

3. The device for improving the personal safety of workers on a construction site according to claim 1, It is characterized in that The judgment model is used to judge whether the human-machine distance is within the safe area, specifically: Assume that the calculation formula of the man-machine distance d is as follows: Among them, r i ,y i and z i are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the worker, respectively. 机 ,y 机 and z 机 They are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the construction machinery; Assume the worker safety distance is L, and set: L=s+x In the formula, s is the dangerous distance of the construction machinery, and x is a preset constant; When d>L, the worker is judged to be in the safe area. When s<d≤L, the worker is judged to have entered the buffer zone and may be in danger. An optimized Markov prediction model is established to judge the direction of the worker at the next moment. When d≤s, the worker is judged to be in the dangerous area.

4. A method for improving the personal safety of workers on construction sites, It is characterized in that include: Through the information collection unit, the distance between the worker and the construction machinery is measured and the worker's location information is obtained, and the data is saved and uploaded to the analysis and processing unit; The analysis and processing unit receives the data uploaded by the information collection unit and performs analysis and processing, wherein the analysis and processing includes: judging whether the distance between the human and the machine is within the safe area through a judgment model; if the worker is found to be in the buffer area, establishing an optimized Markov prediction model to predict the direction of the worker at the next moment; when the prediction model predicts that the worker is about to enter the dangerous area at the next moment, immediately sending an instruction to the early warning unit; if the worker is found to be in the dangerous area, directly sending an instruction to the early warning unit; After receiving the instruction from the analysis and processing unit, the early warning unit turns on the buzzer and the alarm light to remind the worker that he is in the danger zone, and then turns off the buzzer and the alarm light after receiving the instruction from the worker to leave the danger zone; The optimized Markov prediction model is specifically implemented by: constructing a Markov prediction model, improving the Markov model and optimizing parameters; The constructing of the Markov prediction model comprises: Expand the Markov model to a five-element vector group: λ={S,V,π,A,B} Among them, S is the state space, V is the observation symbol space, π is the initial state probability distribution, A is the state transfer matrix, and B is the observation value matrix; The data obtained by the infrared sensor can be used to determine the position of the worker, so that the worker can be reminded in time; set the state space: S={s 1 ,s 2 ,s 3 } s 1 Indicates that the worker is in a safe area and is not in danger at the moment; 2 Indicates that the worker is in the buffer zone and may step into the danger zone; 3 Indicates that the worker is within the danger zone of a large machine and needs to be alerted immediately; Set up the node state sequence as: Q = {q 1 ,q 2 , ..., q r }, q 1 ,q 2 , ..., q r ∈S, number of states N = 3; The danger level of the worker’s current position is determined by the distance between the worker and the machine, and an observation symbol space is set up: V = {v 1 , v 2 , v 3 , v 4 , v 5 },v 1 , v 2 , v 3 , v 4 and v 5 Indicates the degree of danger of the worker's position, the number of observations M = 5, the larger the subscript value, the more dangerous the worker's position; Set up an observation sequence: O = {o 1 , o 2 ,... r },o 1 , o 2 ,... r ∈V; a single observation value represents the location information of the worker observed within a specified time period; Establish the initial state probability distribution: π = {π 1 , π 2 , π 3 }, π i Indicates that the worker's initial state is s i The probability π i =P(q 1 =s 1 ), P is the probability calculation symbol, ∑π i =1(1≤i≤N); Establish the state transfer matrix: A = [a ij ] N×N , a ij Indicates that the worker is from state s i Transfer to j The probability of ij =P(q t+1 =s j |q t =s i ),∑a ij =1(1≤j≤N); Establish the observation matrix: B = [b i (v p )] N×M , b i (v p ) indicates that the worker position is s i The observed value is v p The probability b i (v p )=P(o t =v p |q t =s i ), and ∑b i (v p )=1(1≤p≤M); Based on the forward-backward algorithm, the Markov model parameters are updated to maximize P(O|λ), where Specific: p * =f t (i) Where, 1≤i, j≤M, T is time, f t () is the forward probability of the forward-backward algorithm at the i-th moment, ε t () The backward probability of the forward-backward algorithm at time t; Among them, 1≤i≤M, 1≤k≤N; After updating the parameters using the above method, we get P(O|λ * )>P(O|λ), perform multiple iterations to obtain the maximum likelihood estimate of the Markov model; The Markov model is improved and parameters are optimized, specifically, the HMM-SA-BW algorithm is improved based on the Markov model based on the simulated annealing algorithm and the Baum Welch algorithm to maximize P(O|λ), and the method includes: Initialize model parameters: π, A, B, annealing cooling temperature T 0 , temperature cooling coefficient k, termination temperature T * ; Set the convergence condition ψ=-10 -2 , set the cooling function: T m+1 =kT m , where 0<k<1, T m represents the temperature at time m; Generate M×N independent random variables x that satisfy the normal distribution. The expectation of the normal distribution is 0 and the variance is Let a ik * =a ik +x, 1≤i≤M, 1≤k≤N, when a ik * When a is less than 0, let a ik * =0, the unit is normalized to obtain Update the model and introduce the Momentum first-order optimization acceleration algorithm: Among them, r t is the momentum variable, γ is the momentum hyperparameter, 0≤γ≤1, η t is the learning rate at the tth step, θ is the parameter P(O|λ) that needs to be updated, is the gradient of the loss function with respect to the parameters; Determine whether P(O|λ) satisfies the termination temperature or convergence condition: When P(O|λ) satisfies the condition, P(O|λ) is the optimal solution and the algorithm terminates; otherwise, it returns to the step of initializing the model parameters and continues.

5. The method for improving the personal safety of construction workers according to claim 4, It is characterized in that The information acquisition unit includes an infrared distance measuring sensor, and the infrared distance measuring sensor includes an infrared transmitter and an infrared receiver; The infrared transmitter is used to transmit an infrared signal of a predetermined frequency. The infrared receiver is used to collect the signal reflected after colliding with the obstacle, and perform calculations, and save and upload the data to the analysis and processing unit.

6. The method for improving the personal safety of construction workers according to claim 4, It is characterized in that The judgment model is used to judge whether the human-machine distance is within the safe area, specifically: Assume that the calculation formula of the man-machine distance d is as follows: Among them, r i ,y i and z i are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the worker, respectively. 机 ,y 机 and z 机 They are the real-time three-dimensional x-axis, y-axis, and z-axis coordinates of the construction machinery; Assume the worker safety distance is L, and set: L=s+x In the formula, s is the dangerous distance of the construction machinery, and x is a preset constant; When d>L, the worker is judged to be in the safe area. When s<d≤L, the worker is judged to have entered the buffer zone and may be in danger. An optimized Markov prediction model is established to judge the direction of the worker at the next moment. When d≤s, the worker is judged to be in the dangerous area.

Images