Limited space operation height measuring method and device based on multi-sensor fusion

Abstract

The invention is suitable for the technical field of limited space operation, and provides a limited space operation height measurement method and device based on multi-sensor fusion, and the method comprises the following steps: the device is powered on, and an underground node hardware platform collects the temperature, air pressure value and height value of an underground operation environment, and transmits the values to a ground gateway hardware platform; the ground gateway hardware platform collects the temperature, the air pressure value and the height value of the position where the ground gateway hardware platform is located and uploads sensor data at the two positions to the computing power unit; the computing power unit computes the received data to obtain a final height difference value and transmits the final height difference value back to the ground gateway hardware platform; the ground gateway hardware platform analyzes the height difference value and judges whether the relative working height of the operator is abnormal or not; and if the relative working height of the operator is abnormal, the underground node hardware platform gives an alarm. According to the invention, the position information of the operating personnel can be accurately positioned, and the safety of an operating site in a limited space is effectively guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of confined space operations, and in particular relates to a confined space operation height measurement method and device based on multi-sensor fusion. Background Art

[0002] During construction, workers often enter confined spaces that are semi-enclosed or fully enclosed, isolated from the outside world, and have narrow entrances and exits to carry out construction work. Compared with other working environments, confined spaces are relatively more dangerous, and the control of workers should be stricter.

[0003] At present, the measurement of the working height of personnel in confined spaces mainly relies on traditional measuring tools, or uses a single air pressure altitude sensor to measure the height. Traditional measuring tools require personnel to take measurements and readings, which is not conducive to real-time monitoring of operations. A single air pressure altitude sensor will be affected by the actual conditions of the measurement environment. The air pressure value at any place will change with factors such as time, temperature, wind speed, and season. The longer the operation time, the greater the error in the calculated height. Therefore, how to accurately measure the working height of personnel in confined spaces is of great research significance. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for measuring height of limited space operations based on multi-sensor fusion, aiming to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The limited space operation height measurement device based on multi-sensor fusion includes:

[0007] An underground node hardware platform, the underground node hardware platform includes a sensor module, a power module, an alarm module and a communication module; the sensor module collects the temperature, air pressure and altitude of the underground working environment through sensors; the alarm module is used to trigger an alarm when it detects that the height difference exceeds a preset safety range;

[0008] The ground gateway hardware platform communicates with the underground node hardware platform through the LORA module; it is used to receive data from the underground node hardware platform; the ground gateway hardware platform includes a sensor module, a power module and a communication module; the sensor module collects the temperature, air pressure and altitude values ​​of the location of the ground gateway hardware platform through sensors;

[0009] The computing unit communicates with the ground gateway hardware platform through a wireless 2.4G module. The ground gateway hardware platform uploads the data of the ground gateway hardware platform and the underground node hardware platform to the computing unit. The computing unit receives and processes the data from the ground gateway hardware platform, and transmits the calculated height difference back to the ground gateway hardware platform.

[0010] Furthermore, the process of processing data by the computing unit includes:

[0011] In the data processing phase, a MobileNet-Attention-GRU model is built to standardize the data reported by the ground gateway hardware platform;

[0012] During the training phase, the data and label values ​​are analyzed and learned to obtain a trained model; the label value is the height difference between the underground node hardware platform and the ground gateway hardware platform;

[0013] During the testing phase, the actual sensor data set is input into the trained model, and the data set is detected through the MobileNet-Attention-GRU model to output the estimated height difference.

[0014] Furthermore, the data collected in the data processing stage includes the following information, denoted as X, where X represents the following:

[0015] [T,X1 h ,X2 h ,X3 h ,X4 h ,X5 h ,X6 h ,X1 s ,X2 s ,X3 s ,X4 s ,X5 s ,X6 s ] T ;

[0016] Among them, T is the time of sensor acquisition, X1 h ,X2 h ,X3 h ,X4 h ,X5 h ,X6 h They are the temperature value, humidity value, BMP280 air pressure value, BMP280 altitude value, HP303 air pressure value and HP303 altitude value collected by the ground gateway hardware platform; X1 s ,X2 s ,X3 s ,X4 s ,X5 s,X6 s They are the temperature value, humidity value, BMP280 air pressure value, BMP280 altitude value, HP303 air pressure value and HP303 altitude value collected by the underground node hardware platform respectively; the label value is recorded as Y; the data length of 5s is intercepted as the data length, and the input data is expressed as follows:

[0017]

[0018] Among them, X′ is the model data sent to the network training after the sample uses the time window to divide the data.

[0019] Furthermore, the process of constructing the MobileNet-Attention-GRU model is as follows:

[0020] The MobileNet-Attention-GRU model combines MobileNet-Attention and GRU;

[0021] First, in the MobileNet-Attention part, the input features are dimensioned using 1×1 convolution, followed by BN and h-swish. Then, 3×3 deep convolution is used for feature extraction. Then, the lightweight attention mechanism is used to adjust the weights of the features. Then, 1×1 convolution is used to reduce the feature dimension, followed by BN and h-swish. Finally, the input and output features of the MobileNet-Attention part are fused through the residual structure.

[0022] In the GRU part, there are two structures: reset gate and update gate;

[0023] Finally, the models obtained by MobileNet-Attention and GRU are fused to obtain the MobileNet-Attention-GRU model. The model is sparsely processed using L1 regularization to obtain the estimated height difference. The cost function formula of L1 regularization is as follows:

[0024]

[0025] Among them, w is the parameter of the MobileNet-Attention model, α is the regularization coefficient, L is the loss function term without adding regularization, x is the feature set, and y is the label set.

[0026] Furthermore, the attention mechanism is divided into a coordinate information embedding part and a coordinate attention generation part;

[0027] The coordinate information embedding module uses pooling in two directions for the input sensor features, divides the features into feature maps in two independent directions, learns the sequence relationship of all sensors in a single time through X-average pooling, and learns the sequence relationship between single sensors in multiple times through Y-average pooling, and then uses 1×1 convolution to adjust the channels;

[0028] The coordinate attention generation module fuses the features extracted in the previous step and splits them again, uses 1×1 convolution to adjust different channels, obtains weight information of two different sequences through Sigmoid activation function, and finally multiplies the sequence weight information with the input features to obtain the weight-adjusted features.

[0029] Furthermore, in the training phase, the model was trained using the gradient descent method, using the Adam optimizer with an initial learning rate of 10 -4 , the learning rate is adjusted using cosine annealing decay, the batch size is set to 64, and the training rounds are set to 100000.

[0030] The limited space operation height measurement method based on multi-sensor fusion includes the following steps:

[0031] Step 1: The device is powered on, and the underground node hardware platform collects the temperature, air pressure and altitude values ​​of the underground working environment and sends them to the ground gateway hardware platform;

[0032] Step 2: After receiving the uploaded data from the underground node hardware platform, the ground gateway hardware platform collects the temperature, air pressure and altitude values ​​of the location of the ground gateway hardware platform; the ground gateway hardware platform uploads the sensor data of the underground working environment and the location of the ground gateway hardware platform to the computing unit;

[0033] Step 3: The computing unit calculates the received data, obtains the final height difference and transmits it back to the ground gateway hardware platform;

[0034] Step 4: The ground gateway hardware platform analyzes the height difference to determine whether the relative working height of the operator is abnormal; if the relative working height of the operator is abnormal, the ground gateway hardware platform sends a signal to the underground node hardware platform, and the underground node hardware platform issues an alarm.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention can accurately locate the position information of the operators and effectively ensure the safety of the limited space operation site. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the structure of the device of the present invention.

[0038] Figure 2 This is a framework diagram of the ground gateway hardware platform in the present invention.

[0039] Figure 3 This is a framework diagram of the underground node hardware platform in the present invention.

[0040] Figure 4 Flow chart of the method of the present invention.

[0041] Figure 5 This is a flow chart of the computing unit processing data in the present invention.

[0042] Figure 6 This is the structure diagram of the MobileNet-Attention-GRU model in the present invention.

[0043] Figure 7 This is the MobileNet-Attention structure diagram in the present invention. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0045] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0046] like Figure 1 As shown, a limited space operation height measurement device based on multi-sensor fusion is provided in one embodiment of the present invention, comprising:

[0047] An underground node hardware platform, the underground node hardware platform includes a sensor module, a power module, an alarm module and a communication module; the sensor module collects the temperature, air pressure and altitude of the underground working environment through sensors; the alarm module is used to trigger an alarm when it detects that the height difference exceeds a preset safety range;

[0048] The ground gateway hardware platform communicates with the underground node hardware platform through the LORA module; it is used to receive data from the underground node hardware platform; the ground gateway hardware platform includes a sensor module, a power module and a communication module; the sensor module collects the temperature, air pressure and altitude values ​​of the location of the ground gateway hardware platform through sensors;

[0049] The computing unit communicates with the ground gateway hardware platform through a wireless 2.4G module. The ground gateway hardware platform uploads the data of the ground gateway hardware platform and the underground node hardware platform to the computing unit. The computing unit receives and processes the data from the ground gateway hardware platform, and transmits the calculated height difference back to the ground gateway hardware platform.

[0050] In the embodiment of the present invention, Figure 3 As shown, in the underground node hardware platform, the main control (MCU) is APM32E103RET6, which has functions such as node sensor data collection, node data upload, and control alarm.

[0051] The BMP280 barometric pressure and altitude sensor is connected to the MCU via the UART serial port protocol and can directly read the barometric pressure and altitude values. The HP303B barometric pressure and altitude sensor communicates with the MCU via the I2C protocol to read the barometric pressure and altitude values. The AHT20 temperature and humidity sensor uses the IIC protocol to communicate and read the ambient temperature and humidity values.

[0052] The power supply is a 2000mAh lithium-ion battery, TP4056 is used as the charging chip, and the AMS1117-3.3V regulator converts the voltage to 3.3V to power the entire system.

[0053] The alarm module is mainly composed of LED lights and vibration motors. When the height is violated, the LED light flashes and the motor vibrates.

[0054] The LORA module model is E59-470NW22SC node module, which realizes communication with the gateway.

[0055] like Figure 2 As shown, in the ground gateway hardware platform, the main control (MCU) is APM32F407IGT6, which has the functions of gateway sensor data collection, node data reception, data reporting, and communication module control.

[0056] The sensor module has the same hardware platform as the underground node.

[0057] The power supply is a 18650 rechargeable lithium-ion battery. The battery model is ICR-18650-14.8V / 17.6AH, with a typical capacity of 17500mAh. It has a built-in DC module that can directly output 5V voltage, and convert the voltage to 3.3V through the AMS1117-3.3V regulator to power the entire system.

[0058] The wireless 2.4G module model is Ai-WB2-12F, which supports Wi-Fi 802.11b / g / n protocol and BLE 5.0 protocol, and communicates with the computing unit via Wi-Fi.

[0059] Because the ground has a certain interference and shielding effect on electromagnetic signals, the application scenario of the present invention is the limited underground space, so the LORA communication module is used to achieve the coverage networking requirements of small data in complex environments. The LORA module model is the E59-470NW22SG gateway, which is mainly used to realize the communication function with underground nodes.

[0060] The computing unit uses the Jetson Orin NX 16GB embedded platform, which has 100TOPS AI performance, is equipped with a 1024-core NVIDIA Ampere architecture GPU with 32 Tensor Cores, and a maximum CPU frequency of 2GHz. It supports 3x UART, 2x SPI, 2xI2S, 4x I2C, 1x CAN, DMIC and DSPK, PWM, and GPIO interfaces.

[0061] As a preferred embodiment of the present invention, the present invention transforms the data processing problem of the sensor into a regression problem in supervised learning. The process of processing data by the computing unit includes:

[0062] In the data processing phase, a MobileNet-Attention-GRU model is built to standardize the data reported by the ground gateway hardware platform;

[0063] In the training phase, the data and label values ​​(the height difference between the underground node hardware platform and the ground gateway hardware platform) are analyzed and learned to obtain a trained model;

[0064] In the testing phase, the actual sensor data set is input into the trained model, and the data set is tested through the MobileNet-Attention-GRU model to output the estimated height difference. Figure 5 shown.

[0065] As a preferred embodiment of the present invention, in the data processing stage, since in actual use, the air pressure height is affected by the actual conditions of the measurement environment, the change of air pressure is affected by the ambient temperature and humidity, and also includes the periodic change of air pressure. The so-called periodic change of air pressure refers to the regular fluctuation of air pressure over time. Taking into account the time and environmental information, the collected data contains the following information, denoted as X, where X represents the following:

[0066] [T,X1 h ,X2 h ,X3 h ,X4 h ,X5 h ,X6 h ,X1 s ,X2s ,X3 s ,X4 s ,X5 s ,X6 s ] T ;

[0067] Among them, T is the time of sensor acquisition, X1 h ,X2 h ,X3 h ,X4 h ,X5 h ,X6 h They are the temperature value, humidity value, BMP280 air pressure value, BMP280 altitude value, HP303 air pressure value and HP303 altitude value collected by the ground gateway hardware platform; X1 s ,X2 s ,X3 s ,X4 s ,X5 s ,X6 s They are the temperature value, humidity value, BMP280 air pressure value, BMP280 height value, HP303 air pressure value and HP303 height value collected by the underground node hardware platform respectively; the label value is recorded as Y; the collected data is a sequence vector containing time characteristics. It is impossible to judge the height only by a set of data, and errors will also be caused by collecting wrong values. Therefore, the present invention intercepts the data length of 5s as the data length, which can avoid the problem of abnormal height caused by short-term errors in sensor data. The input data is expressed as follows:

[0068]

[0069] Among them, X′ is the model data sent to the network training after the sample uses the time window to divide the data.

[0070] As a preferred embodiment of the present invention, the data processing algorithm mainly uses deep learning technology to construct a computing model. Convolutional neural network (CNN) is superior to recurrent neural network (RNN) in terms of local features. CNN can extract more features of the data, but it will cause the loss of time series information. RNN is the opposite. RNN and its variant structures have a certain memory function and can process input data of any length. The model does not change with the change of input length, but it is easy to lose early input information and is easily affected by gradient vanishing or gradient exploding. Therefore, the present invention constructs a computing model by integrating CNN and RNN in parallel, namely the MobileNet-Attention-GRU model. The model structure is as follows Figure 6 shown. Figure 6 The original data is the processed data X.

[0071] MobileNet-Attention is a network model improved based on CNN, which is mainly constructed based on MobileNet and the attention mechanism Attention. MobileNet is a lightweight model in CNN. It is a lightweight deep convolutional neural network developed for embedded devices. Its core idea is to use lightweight depth-separable convolution instead of ordinary convolution. The human brain will pay more attention to things of interest. The attention mechanism is a method for obtaining features of the area of ​​interest in the neural network, thereby increasing the weight of key features. MobileNet-Attention is as follows Figure 7 As shown in the figure, by embedding a lightweight channel attention mechanism (Coordinate Attention) containing position information in the depth-separable convolution, more sufficient features of the region of interest can be extracted. First, the input feature is dimensionally increased using a 1×1 convolution, and then a 3×3 depth convolution is used for feature extraction. After that, the lightweight attention mechanism is used to adjust the feature weights. Then, a 1×1 convolution is used to reduce the feature dimension. Finally, the input and output features of the MobileNet-Attention part are fused through a residual structure.

[0072] The present invention uses BN+h-swish to perform normalization after 1×1 convolution and reduce the computational cost caused by deepening the network.

[0073] The attention mechanism is divided into coordinate information embedding and coordinate attention generation.

[0074] The coordinate information embedding module uses pooling in two directions for the input sensor features, divides the features into two independent feature maps, learns the sequence relationship of all sensors in a single time through X-average pooling, and learns the sequence relationship between single sensors in multiple times through Y-average pooling, thereby achieving effective feature extraction. Then, 1×1 convolution is used to adjust the channel;

[0075] The coordinate attention generation module fuses the features extracted in the previous step and splits them again, uses 1×1 convolution to adjust different channels, obtains weight information of two different sequences through Sigmoid activation function, and finally multiplies the sequence weight information with the input features to obtain the weight-adjusted features.

[0076] Compared with LSTM, the GRU model has a simplified internal structure and improved accuracy. GRU uses two structures: reset gate and update gate. The reset gate determines the historical information that needs to be forgotten, while the update gate determines the information that needs to be passed on. Compared with LSTM, GRU has fewer parameters, thus greatly shortening the time spent on network training.

[0077] Finally, the models obtained by MobileNet-Attention and GRU are fused to obtain the MobileNet-Attention-GRU model. At the same time, to prevent overfitting, the model is sparsely processed using L1 regularization to obtain the estimated height difference. The cost function formula of L1 regularization is as follows:

[0078]

[0079] Among them, w is the parameter of the MobileNet-Attention model, α is the regularization coefficient, L is the loss function term without adding regularization, x is the feature set, and y is the label set.

[0080] As a preferred embodiment of the present invention, the training phase specifically includes:

[0081] (1) Training settings

[0082] The data source is the sensor data collected in the field, with a total of 30 hours of data. After data cleaning and screening, there are 106,540 valid data in total, and 5-fold cross validation is used for training.

[0083] (2) Training environment

[0084] The training environment uses the Pytorch deep learning framework in the Visual Studio Code integrated development environment. The hardware configuration is I5 12400F (CPU), Nvidia RTX 2070super 8GB (GPU), and 32GB (memory).

[0085] (3) Evaluation indicators

[0086] The root mean square error (RMSE) is used as the evaluation index. The definition of RMSE is as follows:

[0087]

[0088] Among them, ε RMSE Used to represent the final root mean square error, M is the number of samples, y i and They represent the true value and predicted value of the i-th sample respectively.

[0089] (4) Training methods

[0090] The experiment adopts the gradient descent method for training, uses the Adam (Adaptive Momentum Estimation) optimizer, and the initial learning rate is 10 -4 , use cosine annealing decay to adjust the learning rate, set the batch size to 64, set the training rounds to 100000, and use the early stopping method to prevent the network from overfitting.

[0091] As a preferred embodiment of the present invention, in the testing phase, the hardware environment deployed in the testing phase is a Jetson Orin NX 16GB embedded platform, and the software environment is consistent with that in the training phase.

[0092] like Figure 4 As shown, a method for measuring height of limited space operation based on multi-sensor fusion is provided in one embodiment of the present invention, comprising the following steps:

[0093] Step 1: The device is powered on, and the underground node hardware platform collects the temperature, air pressure and altitude values ​​of the underground working environment and sends them to the ground gateway hardware platform;

[0094] Step 2: After receiving the uploaded data from the underground node hardware platform, the ground gateway hardware platform collects the temperature, air pressure and altitude values ​​of the location of the ground gateway hardware platform; the ground gateway hardware platform uploads the sensor data of the underground working environment and the location of the ground gateway hardware platform to the computing unit;

[0095] Step 3: The computing unit calculates the received data, obtains the final height difference and transmits it back to the ground gateway hardware platform;

[0096] Step 4: The ground gateway hardware platform analyzes the height difference to determine whether the relative working height of the operator is abnormal; if the relative working height of the operator is abnormal, the ground gateway hardware platform sends a signal to the underground node hardware platform, and the underground node hardware platform issues an alarm.

[0097] The above are only preferred embodiments of the present invention. It should be pointed out that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These should also be regarded as the protection scope of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A limited space operation height measurement device based on multi-sensor fusion, characterized in that: include: An underground node hardware platform, the underground node hardware platform comprising a sensor module, a power module, an alarm module and a communication module; The sensor module collects the temperature, air pressure and altitude values ​​of the underground working environment through sensors; The alarm module is used to trigger an alarm when it detects that the height difference exceeds a preset safety range; The ground gateway hardware platform communicates with the underground node hardware platform through the LORA module; Used to receive data from the underground node hardware platform; the ground gateway hardware platform includes a sensor module, a power module and a communication module; the sensor module collects the temperature, air pressure and altitude values ​​of the location of the ground gateway hardware platform through sensors; The computing unit communicates with the ground gateway hardware platform through a wireless 2.4G module. The ground gateway hardware platform uploads the data of the ground gateway hardware platform and the underground node hardware platform to the computing unit. The computing unit receives and processes the data from the ground gateway hardware platform, and transmits the calculated height difference back to the ground gateway hardware platform.

2. The limited space operation height measurement device based on multi-sensor fusion according to claim 1 is characterized in that: The process of processing data by the computing unit includes: In the data processing phase, a MobileNet-Attention-GRU model is built to standardize the data reported by the ground gateway hardware platform; During the training phase, the data and label values ​​are analyzed and learned to obtain a trained model; the label value is the height difference between the underground node hardware platform and the ground gateway hardware platform; During the testing phase, the actual sensor data set is input into the trained model, and the data set is detected through the MobileNet-Attention-GRU model to output the estimated height difference.

3. The limited space operation height measurement device based on multi-sensor fusion according to claim 2 is characterized in that: The data collected in the data processing stage contains the following information, denoted as X, where X represents the following: [T,X1 h ,X2 h ,X3 h ,X4 h ,X5 h ,X6 h ,X1 s ,X2 s ,X3 s ,X4 s ,X5 s ,X6 s ] T ; Among them, T is the time of sensor acquisition, X1 h ,X2 h ,X3 h ,X4 h ,X5 h ,X6 h They are the temperature value, humidity value, BMP280 air pressure value, BMP280 altitude value, HP303 air pressure value and HP303 altitude value collected by the ground gateway hardware platform; X1 s ,X2 s ,X3 s ,X4 s ,X5 s ,X6 s They are the temperature value, humidity value, BMP280 air pressure value, BMP280 altitude value, HP303 air pressure value and HP303 altitude value collected by the underground node hardware platform respectively; the label value is recorded as Y; the data length of 5s is intercepted as the data length, and the input data is expressed as follows: Among them, X′ is the model data sent to the network training after the sample uses the time window to divide the data.

4. According to the device for measuring height of limited space operation based on multi-sensor fusion according to claim 1, the process of constructing the MobileNet-Attention-GRU model is as follows: The MobileNet-Attention-GRU model combines MobileNet-Attention and GRU; First, in the MobileNet-Attention part, the input features are dimensioned using 1×1 convolution, followed by BN and h-swish. Then, 3×3 deep convolution is used for feature extraction. Then, the lightweight attention mechanism is used to adjust the weights of the features. Then, 1×1 convolution is used to reduce the feature dimension, followed by BN and h-swish. Finally, the input and output features of the MobileNet-Attention part are fused through the residual structure. In the GRU part, there are two structures: reset gate and update gate; Finally, the models obtained by MobileNet-Attention and GRU are fused to obtain the MobileNet-Attention-GRU model. The model is sparsely processed using L1 regularization to obtain the estimated height difference. The cost function formula of L1 regularization is as follows: in, w is the parameter of the MobileNet-Attention model, α is the regularization coefficient, L is the loss function term without adding regularization, x is the feature set, and y is the label set.

5. According to the device for measuring height of limited space operation based on multi-sensor fusion according to claim 1, the attention mechanism is divided into a coordinate information embedding part and a coordinate attention generation part; The coordinate information embedding module uses pooling in two directions for the input sensor features, divides the features into feature maps in two independent directions, learns the sequence relationship of all sensors in a single time through X-average pooling, and learns the sequence relationship between single sensors in multiple times through Y-average pooling, and then uses 1×1 convolution to adjust the channels; The coordinate attention generation module fuses the features extracted in the previous step and splits them again, uses 1×1 convolution to adjust different channels, obtains weight information of two different sequences through Sigmoid activation function, and finally multiplies the sequence weight information with the input features to obtain the weight-adjusted features.

6. According to the device for measuring height of limited space operation based on multi-sensor fusion in claim 1, in the training phase, the model is trained by gradient descent method, using Adam optimizer, and the initial learning rate is 10 -4 , the learning rate is adjusted using cosine annealing decay, the batch size is set to 64, and the training rounds are set to 100000.

7. A method for measuring height of limited space operations based on multi-sensor fusion, characterized in that: The following steps are involved: Step 1: The device is powered on, and the underground node hardware platform collects the temperature, air pressure and altitude values ​​of the underground working environment and sends them to the ground gateway hardware platform; Step 2: After receiving the uploaded data from the underground node hardware platform, the ground gateway hardware platform collects the temperature, air pressure and altitude values ​​of the location of the ground gateway hardware platform; The ground gateway hardware platform uploads sensor data from both the underground working environment and the location of the ground gateway hardware platform to the computing unit; Step 3: The computing unit calculates the received data, obtains the final height difference and transmits it back to the ground gateway hardware platform; Step 4: The ground gateway hardware platform analyzes the height difference to determine whether the relative working height of the operator is abnormal; if the relative working height of the operator is abnormal, the ground gateway hardware platform sends a signal to the underground node hardware platform, and the underground node hardware platform issues an alarm.

Images