A long-distance wireless sensing method for unstructured environments in coal mines

Abstract

The present invention discloses a long-distance wireless sensing method for the unstructured environment of coal mines. The method comprises the following steps: an intelligent detection robot is equipped with a directional transmitting antenna and a pair of receiving antennas arranged side by side on the left and right sides; directional transmitting antennas facing the same direction are installed at set intervals along the length direction of an underground coal mine tunnel; and a reflector is installed at the optimal reflection installation position on the tunnel wall of the underground coal mine tunnel. The method uses the transmitting antenna and receiving antenna working periodically for detection. After a disaster occurs, the method can determine the specific location of the trapped person based on the amplitude change direction of the gesture signal of the trapped person and the position of the signal source. The method combines the smoothed MUSIC algorithm and the ISAR algorithm to accurately separate and analyze the reflected signal, thereby determining the number of people in the environment and their relative positions. The method can ensure effective signal transmission and detection in a complex underground coal mine tunnel environment. The method is particularly suitable for rescue detection of trapped people in a post-disaster underground coal mine environment.

Description

Technical Field

[0001] The present invention relates to a long-distance wireless sensing method, in particular to a long-distance wireless sensing method for coal mine non-structural environments that can achieve long-distance personnel positioning and number detection in complex underground coal mine environments, and belongs to the technical field of coal mine safety production. Background Art

[0002] As my country's primary energy source, coal has long provided a solid foundation for stable and rapid economic growth. However, coal mining involves complex underground mining environments and intricate procedures. On the one hand, the underground environment in coal mines is complex and volatile, subject to long-term harsh conditions such as high temperature, high humidity, dust, and gas, and the complex tunnel structures present numerous safety hazards. On the other hand, the risk of coal mine disasters and accidents is gradually increasing, accompanied by various uncertainties in the coal mining process, as coal mining intensity continues to increase and mining depth continues to expand. In post-disaster emergency rescue in underground coal mines, traditional personnel detection methods mostly rely on contact sensors or video surveillance systems. These methods have the following major problems: contact sensors need to be widely deployed underground, which requires a lot of installation and maintenance work. In addition, the sensors are easily affected by environmental factors such as humidity and dust, resulting in reduced detection accuracy. Their detection range is also limited, and they can only effectively detect near the sensor deployment point, making long-distance detection difficult to achieve. Video surveillance systems require a large number of cameras and video transmission equipment, with high initial investment and maintenance costs. In addition, video signals are easily interfered with by factors such as light and dust underground, resulting in reduced image quality. There is also the problem of protecting personnel privacy.

[0003] To achieve effective rescue detection in post-disaster underground coal mine environments, industry researchers have begun exploring non-contact detection technologies. Wireless sensing and detection methods based on radio signals have gradually attracted attention due to their advantages such as easy installation, low cost, and strong adaptability. Among the many radio technologies, LoRa technology has become an ideal choice for wireless sensing and detection in underground environments due to its low power consumption, long-distance transmission, and strong penetration. However, wireless sensing and detection technology based on LoRa signals still faces many challenges in complex underground environments. For example, signal stability is significantly affected by the narrow and tortuous underground tunnels, reflections, and attenuation. Wireless sensing accuracy is poor in identifying and locating the specific location and status of trapped personnel. How to ensure effective signal transmission and detection in the unstructured environment of underground coal mines with complex corners and obstacles remains a difficult problem that the industry urgently needs to solve. Summary of the Invention

[0004] In response to the problems existing in the above-mentioned existing technologies, the present invention provides a long-distance wireless sensing method for the non-structured environment of coal mines, which can effectively sense, accurately separate and analyze reflected signals in the non-structured environment and obstacle environment underground in coal mines, and thus obtain accurate information on the number of people. It is particularly suitable for rescue detection of trapped people in the underground environment of coal mines after disasters.

[0005] To achieve the above-mentioned purpose, the non-contact advanced detection system used in the long-distance wireless sensing method for the unstructured environment of coal mines includes an intelligent detection robot, a receiving antenna, a transmitting antenna and a reflector. A pair of receiving antennas are installed side by side on the intelligent detection robot, and the transmitting antenna is a directional transmitting antenna, and the transmitting antenna is set to multiple, each transmitting antenna is equipped with a LoRa node, one transmitting antenna is installed on the intelligent detection robot, and the remaining transmitting antennas are installed in the underground coal mine tunnel. The reflector with a metal sheet structure is installed on the wall of the underground coal mine tunnel. The autonomously movable intelligent detection robot includes a signal processing module, and the signal processing module is electrically connected to a pair of receiving antennas and a transmitting antenna on the intelligent detection robot respectively;

[0006] The long-distance wireless sensing method for unstructured coal mine environments specifically includes the following steps:

[0007] Step 1, Equipment Arrangement: Install transmitting antennas at set intervals along the length of the coal mine tunnel, with the transmitting antennas facing the same direction. Set the transmission cycle and transmission time of each transmitting antenna. After the transmitting antennas are installed, record the installation position and specific operating time of each transmitting antenna. After determining the optimal reflective installation position of the reflector, install the reflector on the wall of the coal mine tunnel. Conduct on-site exploration in the coal mine tunnel to determine the location with good signal detection effect. Mark and record the specific coordinates and environmental description of each location with good signal detection effect.

[0008] Step 2, post-disaster detection: After a disaster occurs, the coal mine tunnel layout data with locations with good detection results marked is first input into the intelligent detection robot. The intelligent detection robot is then controlled to enter the coal mine tunnel and go to the marked locations with good detection results to perform signal detection. The intelligent detection robot waits for at least one complete signal transmission cycle at each good detection result location to receive signals from all transmitting antennas.

[0009] Step2-1. Determine the location of the trapped personnel: The signal processing module records and analyzes the received signals to determine whether there is a gesture signal among them. When a gesture signal is detected, the signal processing module first determines the source of the gesture signal. If the source is a certain transmitting antenna in the coal mine roadway, the signal processing module determines that the trapped personnel are near the transmitting antenna. If the source is the transmitting antenna on the intelligent detection robot, the signal processing module first judges the amplitude trend of the gesture signal, and then determines on which side of the intersection the trapped personnel are located relative to the intelligent detection robot according to the amplitude fluctuation direction of the gesture signal.

[0010] Step2-2. Determine the number of trapped personnel: After confirming the existence of trapped personnel, process the reflected signals by eliminating the static signals and direct path signals in the received signals, separate the reflected signals in different directions, determine the arrival angle of the reflected signals, and calculate the position of the trapped personnel relative to the receiving antenna and the number of trapped personnel.

[0011] Further, in Step2-1, when the signal processing module determines on which side of the intersection the trapped personnel are located relative to the intelligent detection robot, an upward amplitude fluctuation direction of the gesture signal indicates that the trapped personnel are at the right corner of the intersection, and a downward amplitude fluctuation direction of the gesture signal indicates that the trapped personnel are at the left corner of the intersection. If there are both upward and downward amplitude fluctuation directions, it means that there are trapped personnel at both the left and right corners of the intersection.

[0012] Further, in Step2-1, when the signal processing module judges the amplitude trend of the gesture signal, it uses the singular spectrum analysis method, multi-scale sliding window fluctuation detection method, and average gradient judgment method for judgment.

[0013] The specific steps of the singular spectrum analysis method are as follows:

[0014] ① Embed the one-dimensional time series into a high-dimensional space. Given a time series (X1, X2, …, X N ), select a window length L (usually L < N / 2), and construct a trajectory matrix X of L×KL, where K = N - L + 1. Each column of the trajectory matrix is a subsequence of the time series, expressed as follows:

[0015] X = [X1, X2, …, X K , X I = (x i , x i+1 , …, x i+L-1 ) T

[0016] Perform SVD decomposition on the trajectory matrix X to obtain:

[0017]

[0018] Where: f is a nonlinear function; λ i It's XX T The characteristic value of U I and Vi are the left and right singular vectors respectively; s is the number of non-zero singular values; αX0 is an additional matrix that increases the degree of freedom;

[0019] ② According to the size and physical meaning of the eigenvalue, the decomposed components are grouped. The larger eigenvalue corresponds to the trend and periodic components, and the smaller eigenvalue corresponds to the noise component.

[0020] ③ Perform inverse transformation on the selected components to reconstruct the time series. By selecting different components, the trend, cycle and noise parts of the time series can be extracted respectively;

[0021] The specific steps of the multi-scale sliding window fluctuation detection method are as follows:

[0022] ①Select sliding windows of different sizes for data processing;

[0023] ② Calculate the degree of fluctuation of the data in each window. If the fluctuation in the window is less than the predetermined threshold, the data in the window is processed as 0;

[0024] ③ Perform fluctuation detection on each sliding window size separately, as follows:

[0025] a. Use the maximum window size to process the entire data and obtain the first-level processing results;

[0026] b. Process the first-level processing results using a medium window size to obtain the second-level processing results;

[0027] c. Process the second-level processing results using the minimum window size to obtain the final result;

[0028] The specific steps of the average gradient judgment method are as follows:

[0029] ① Calculate the gradient of the initial part of each gesture signal amplitude data. The gradient formula is:

[0030]

[0031] Where: ΔA is the amplitude difference between adjacent sampling points; Δt is the sampling interval;

[0032] ②Average the gradient values ​​of each segment of data to obtain the average gradient of the segment of data:

[0033]

[0034] Where: Ns is the number of sampling points of this segment of data;

[0035] ③ Judge the trend of the signal based on the sign and size of the average gradient. If the average gradient is positive, it means that the signal amplitude is increasing; if the average gradient is negative, it means that the signal amplitude is decreasing.

[0036] Furthermore, in Step 2-2, when processing the reflected signal by eliminating the static signal and the direct path signal in the received signal, the inverse synthetic aperture radar algorithm and the smoothed MUSIC algorithm are applied to perform in-depth analysis of the signal;

[0037] When applying the inverse synthetic aperture radar algorithm to perform in-depth signal analysis, the trajectory formed by the human gesture movement is regarded as the antenna array. The spatial direction angle calculation formula is as follows:

[0038]

[0039] Where: Angle[θ,n] is the signal function along the spatial direction θ at the measurement time n; λ is the wavelength; Δd is the spatial distance between consecutive antennas in the array; S[n+i] is the antenna array;

[0040] Estimate the spatial distance Δd between consecutive antennas in the simulation array. Δd is expressed as Δd = vT, where T is the sampling period and v is the velocity. The default value of v is v = 1 m / s.

[0041] When applying the smoothed MUSIC algorithm to perform in-depth analysis of a signal, the specific steps are as follows:

[0042] ① Construct the covariance matrix R of the received signal, R=E[xx H ];

[0043] ② Calculate the eigenvalues ​​and eigenvectors of the covariance matrix and divide it into signal subspace and noise subspace;

[0044] ③By processing the eigenvectors of the noise subspace, the MUSIC space spectrum function is formed:

[0045]

[0046] Where: W i Is a transformation matrix used to filter the signal; e i is the eigenvector of the noise subspace; a(θ) is the array manifold vector;

[0047] ④ By finding the peak of the MUSIC spectrum, the arrival direction of the signal is estimated, and finally a clear trajectory formed by human movement is obtained in the time direction spectrum. The number of people in the environment is obtained based on the number of trajectories.

[0048] Furthermore, in Step 2, when the intelligent detection robot goes to the marked locations with good detection effects to perform signal detection, it selects at least two locations with good detection effects at each intersection to perform signal detection.

[0049] Furthermore, in Step 2-1, if the emission source is a transmitting antenna on the intelligent detection robot, the detection is repeated three times at the location with good detection effect, and the case where the detection result is greater than or equal to twice is selected.

[0050] Furthermore, in Step 1, when installing transmitting antennas in underground tunnels of coal mines, for intersections and T-junctions, according to the detection range of the receiving antenna on the intelligent detection robot, the geometric center of the intersection is used as the range center, and the transmitting antenna is not installed within the set range.

[0051] Furthermore, in Step 1, when determining the optimal reflective installation position of the reflector, detection equipment is used to conduct on-site exploration of the underground tunnels of the coal mine, and the optimal reflective installation position of the reflector is determined through experiments and data analysis.

[0052] In Step 1, when determining locations with good signal detection effects, for crossroads and T-junctions, multiple locations with good detection effects are determined for each intersection.

[0053] Compared with existing technologies, this long-distance wireless sensing method for unstructured coal mine environments uses an intelligent detection robot equipped with a directional transmitting antenna and a pair of receiving antennas arranged side by side. Directional transmitting antennas facing the same direction are installed at set intervals along the length of the coal mine tunnel. Each transmitting antenna is equipped with a LoRa node, and reflectors are installed at the optimal reflective installation position on the tunnel wall of the coal mine. The transmitting and receiving antennas work periodically to avoid signal interference and ensure the accuracy and reliability of signal reception. After a disaster occurs, the specific location of trapped people can be determined based on the amplitude change direction and signal source location of their gesture signals. The smoothed MUSIC algorithm and the ISAR algorithm are combined to accurately separate and analyze the reflected signals, and then determine the number and relative positions of people in the environment, thereby achieving long-distance perception and accurate positioning of trapped people. It can ensure effective signal transmission and detection in complex coal mine tunnel environments, and is particularly suitable for rescue detection and locating trapped people in coal mine environments after disasters. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 is a flow chart of the present invention;

[0055] Figure 2 This is a schematic diagram of the gesture signal return receiving antenna of the present invention;

[0056] Figure 3Schematic diagram of the reflection area of ​​different receiving antennas returned by the present invention;

[0057] Figure 4 This is a schematic diagram of the placement of the transmitting antenna and reflector of the present invention;

[0058] Figure 5 This is a schematic diagram of the sensing range of a T-junction of the present invention;

[0059] Figure 6 This is a schematic diagram of an antenna array for simulating human motion using ISAR in the present invention. DETAILED DESCRIPTION

[0060] The present invention will be further described below with reference to the accompanying drawings.

[0061] The non-contact advanced detection system used in this long-distance wireless sensing method for the non-structural environment of coal mines includes an intelligent detection robot, a receiving antenna, a transmitting antenna and a reflector. A pair of receiving antennas are installed side by side on the intelligent detection robot. The transmitting antenna is a directional transmitting antenna, and there are multiple transmitting antennas. Each transmitting antenna is equipped with a LoRa node. One transmitting antenna is installed on the intelligent detection robot, and the remaining transmitting antennas are installed in the underground coal mine tunnel. The reflector with a metal sheet structure is installed on the wall of the underground coal mine tunnel. The autonomously movable intelligent detection robot includes a signal processing module, and the signal processing module is electrically connected to a pair of receiving antennas and a transmitting antenna on the intelligent detection robot respectively.

[0062] like Figure 1 As shown, the long-distance wireless sensing method for the unstructured environment of coal mines specifically includes the following steps:

[0063] Step 1, equipment layout:

[0064] Step 1-1: Install transmitting antennas at set intervals along the length of the underground coal mine tunnel, and the transmitting antennas should face the same direction. To reduce signal interference and lower costs, for crossroads and T-junctions, based on the detection range of the receiving antenna on the intelligent detection robot, the geometric center of the intersection is used as the range center, and no transmitting antenna is installed within the set range. After the transmitting antenna is installed, the position of each transmitting antenna is recorded.

[0065] For example, the transmitting antenna is a 9dbi directional transmitting antenna, the receiving antenna model on the intelligent detection robot is USRPX310, the carrier frequency of the LoRa node transmitting signal is 915MHz, the linear frequency modulation bandwidth BW is 125KHz, the spreading factor SF is 12, the coding rate CR is 4 / 8, and the sampling rate is 500Hz. Figure 4 、 Figure 5As shown, a transmitting antenna is installed every 500 meters along the length of each lane. For crossroads and T-junctions, the receiving antenna on the intelligent detection robot can detect signals within 150 meters of the corner. Therefore, no transmitting antenna is installed within 100 meters of each intersection to reduce the number of transmitting antennas used and reduce costs.

[0066] Step 1-2: To ensure that the transmitting antennas do not interfere with each other, set the transmission cycle and transmission time of each transmitting antenna to avoid signal interference. The transmitting antennas work periodically, and transmitting antennas that may cause interference will not work at the same time. Record the location and specific operating time of each transmitting antenna in detail so that the signal source can be accurately matched during subsequent signal processing.

[0067] Step 1-3, use detection equipment to conduct on-site exploration of the underground tunnels of the coal mine, determine the optimal reflective installation position of the reflector, and then install the reflector on the wall of the underground tunnel of the coal mine. The optimal reflective installation position of the reflector can be determined through experiments and data analysis to ensure that the signal strength and detection effect are enhanced to the greatest extent. When installing the reflector, ensure that the sheet is flat and firm, and try to avoid being blocked by other objects. After the reflector is installed, use the detection equipment again to detect the signal at the installation position of the reflector to verify the effect of the iron sheet on signal enhancement, record and analyze the detection data, and ensure that the detection distance and signal strength can be significantly improved.

[0068] Steps 1-4 involve conducting on-site surveys within the mine tunnels to identify locations with good signal detection. The specific coordinates and environmental descriptions of each location are marked and recorded so that these locations can be quickly found and utilized during the actual rescue process. Multiple locations with good detection performance are identified at each intersection to avoid loss of detection during severe damage. These locations require good signal reception to provide reliable detection results even in the presence of debris and other obstacles after a disaster.

[0069] Step 2, post-disaster detection: After the disaster occurs, the coal mine tunnel layout data marked with good detection effect positions are first input into the intelligent detection robot, and then the intelligent detection robot is controlled to enter the underground coal mine tunnel and go to the marked good detection effect positions for signal detection. At least two good detection effect positions are selected at each intersection for signal detection. Since the transmission time of each transmitting antenna is different, the intelligent detection robot waits for at least one complete signal transmission cycle at each good detection effect position to ensure that it can receive the signals sent by all transmitting antennas and ensure the comprehensiveness and accuracy of the data.

[0070] Step 2-1, determine the location of the trapped person:

[0071] The signal processing module records and analyzes the received signals to determine whether there are gesture signals among them. When a gesture signal is detected, the signal processing module first determines the emission source of the gesture signal. If the emission source is the transmitting antenna in the coal mine roadway (i.e., the trapped person is in the roadway where the transmitting antenna is located, and the intelligent detection robot can directly receive the signal transmitted by the transmitting antenna and reflected by the trapped person's body), then the trapped person is near the transmitting antenna. If the emission source is the transmitting antenna on the intelligent detection robot (i.e., the trapped person is at the intersection position with a signal reflection area, and the intelligent detection robot receives the reflected signal transmitted by the transmitting antenna on the intelligent detection robot and reflected by the trapped person's body), first use singular spectrum analysis (SSA), multi-scale sliding window fluctuation detection, and average gradient judgment method to judge the amplitude trend of the gesture signal, and then judge on which side of the intelligent detection robot the trapped person is located according to the amplitude fluctuation direction of the gesture signal.

[0072] SSA is a non-parametric dimensionality reduction method based on time series, mainly used for signal processing, time series decomposition, and trend analysis. Its core idea is to decompose the time series into a set of independent components (such as trends, cycles, noise) to reveal the hidden structure in the data. The basic steps of SSA are as follows:

[0073] ① Embed the one-dimensional time series into a high-dimensional space. Given a time series (X1, X2, …, X N ), select a window length L (usually L < N / 2), and construct an L×KL trajectory matrix X, where K = N - L + 1. Each column of the trajectory matrix is a subsequence of the time series, expressed as follows:

[0074] X = [X1, X2, …, X K , X I = (x i , x i+1 , …, x i+L-1 ) T

[0075] Perform SVD decomposition on the trajectory matrix X to obtain:

[0076]

[0077] In the formula: f is a non-linear function; λ i is the eigenvalue of XX T ; U I and Vi are the left singular vector and the right singular vector respectively; s is the number of non-zero singular values; αX0 is an additional matrix to increase the degree of freedom.

[0078] ② Group the decomposed components according to the size and physical meaning of the eigenvalues. Generally, larger eigenvalues ​​correspond to trend and cycle components, while smaller eigenvalues ​​correspond to noise components.

[0079] ③ Perform inverse transformation on the selected components to reconstruct the time series. By selecting different components, the trend, cycle and noise parts of the time series can be extracted respectively.

[0080] Multi-scale sliding window fluctuation detection is a method that processes data using sliding windows of different sizes. It can process the parts of the signal with subtle fluctuations to zero, thereby highlighting the significant fluctuations in the signal. This method can be divided into the following steps:

[0081] ① Select sliding windows of different sizes for data processing. The sizes of sliding windows should range from large to small, and three or more windows of different sizes are usually selected.

[0082] ② Calculate the degree of fluctuation for the data within each window. Common fluctuation detection methods include mean difference, variance, standard deviation, etc. If the fluctuation within a window is less than a predetermined threshold, the data in that window is treated as zero.

[0083] ③ Perform fluctuation detection on each sliding window size. The specific method is as follows:

[0084] a. Use the maximum window size to process the entire data and obtain the first-level processing results.

[0085] b. Use a medium window size to process the first-level processing results to obtain the second-level processing results.

[0086] c. Use the minimum window size to process the second-level processing results to obtain the final result.

[0087] The average gradient judgment method is a technique that determines the signal trend by calculating the gradient of the signal amplitude. The specific steps are as follows:

[0088] ① Calculate the gradient of the initial part of each gesture signal amplitude data, that is, the difference between the amplitudes of adjacent sampling points. The gradient formula is:

[0089]

[0090] Where: ΔA is the amplitude difference between adjacent sampling points; Δt is the sampling interval.

[0091] ②Average the gradient values ​​of each segment of data to obtain the average gradient of the segment of data.

[0092]

[0093] Where: Ns is the number of sampling points of this segment of data.

[0094] ③ Judge the trend of the signal based on the sign and size of the average gradient. If the average gradient is positive, it means that the signal amplitude is increasing; if the average gradient is negative, it means that the signal amplitude is decreasing.

[0095] When judging the direction of the trapped person based on the amplitude fluctuation direction of the hand signal, if Figure 2 As shown in the figure, the gray part is the reflected signal when the trapped person's hand does not move, and the green part is the reflected signal caused by the movement of the trapped person's hand. Part of the reflected signal is received by the receiving antenna RX1 on the left (the blue part in the figure), and the other part is received by the receiving antenna RX2 on the right (the orange part in the figure). Since the two receiving antennas installed side by side on the left and right are close to each other, the signals of the blue part and the orange part are also close, so the amplitude change caused by path attenuation should be smaller than the change caused by the different reflection areas. Since the transmitting antenna is a directional transmitting antenna, the signal it sends is conical, and the signal reflected from the wall forms an ellipse that is close to a perfect circle (the signal reflected from the air can be regarded as reflected by an invisible wall at a certain point, and there are multiple such ellipses in the entire environment), so the size of the reflection area is related to the diameter of the ellipse. As shown Figure 3 As shown, x and y are used to represent the diameter of the ellipse of the signal reflected by the wall. According to the geometric relationship, the diameter of the ellipse can be calculated by the following formula:

[0096]

[0097] Where: d is the distance from the emission point to the wall; θ1, θ2, θ3 and θ4 are the incident angle and reflection angle of the signal.

[0098] According to the law of reflection, when the signal is reflected from the right corner, θ∈(0°~90°), and θ1<θ2<θ3<θ4. Since the left and right sides are symmetrical, we will use the right side as an example for analysis.

[0099] When θ∈(0°~90°), The first derivative of Second-order derivative therefore is a concave function, that is, as θ gradually increases, The reduction is getting smaller and smaller.

[0100] because The decrease is getting smaller and smaller. As θ increases, it can be considered that there is a higher probability that x>y at θ∈(0°~90°).

[0101] In order to discuss the probability of x>y, the Monte Carlo method was used for numerical simulation. A large number of randomly generated θ1, θ2, θ3, and θ4 were used for difference comparison. After 1 million experiments, when θ∈(0°~90°), the probability that x is greater than y is about 85%.

[0102] In a non-line-of-sight environment, the trapped person's hand movement reflects more signal toward the receiving antenna. When the signal arrives from the right side of the receiving antenna, the signal strength increase at receiving antenna RX1 is greater than the signal strength increase at receiving antenna RX2 because the signal entering the receiving antenna RX1 from the wall has a larger reflection area.

[0103] Usually when processing LoRa signals, receiving antennas RX1 and RX2 are used to eliminate CFO (Carrier Frequency Offset) and CSO (Sampling Frequency Offset). The resulting amplitude can be expressed as:

[0104]

[0105] Where: amp is the collected amplitude; abs represents the absolute value function; and These are the signals received by the two receiving antennas on the intelligent detection robot.

[0106] Therefore, when the signal is reflected from the right corner, there is about an 85% probability that The increase is greater than The increase in the amount of , which causes the amp to become larger; when the signal is reflected from the left corner, there is about an 85% probability that The increase is less than The increase in amp causes the amp to decrease.

[0107] To improve detection accuracy, during intersection detection, if a person is detected as potentially trapped on either side of the intersection, the detection is repeated twice. Of the three detections, only those with two or more detection results are selected. Extensive experiments have shown that this approach can increase detection accuracy to over 93%.

[0108] That is to say, when the transmitting source is the transmitting antenna on the intelligent detection robot, the upward fluctuation direction of the amplitude of the gesture signal indicates that the trapped person is at the right corner of the intersection, and the downward fluctuation direction of the amplitude of the gesture signal indicates that the trapped person is at the left corner of the intersection. If there are both upward and downward amplitude fluctuation directions at the same time, it means that there are trapped people at both the left and right corners of the intersection.

[0109] Step 2-2, determine the number of trapped people:

[0110] After confirming the presence of trapped people, the system eliminates static signals and direct path signals from the received signal, further processes the reflected signal, and applies the inverse synthetic aperture radar (ISAR) algorithm and smoothed MUSIC algorithm to conduct in-depth analysis of the signal. It separates the reflected signals in different directions and determines the arrival angle of the reflected signal. It calculates the position of the trapped people relative to the receiving antenna and the number of trapped people, so that rescuers can carry out precise rescue.

[0111] ISAR is a technology that exploits the motion of a target to simulate an antenna array. In a traditional antenna array, multiple antennas simultaneously receive signals from a target and process this information to determine the target's orientation (θ) relative to the array. In ISAR, however, there is only one receiving antenna. However, because the target is moving, continuous temporal measurements simulate an inverse antenna array—in other words, a moving human body becomes an antenna array. By processing these continuous measurements, the spatial orientation of the human body can be determined using standard antenna array beam steering.

[0112] like Figure 6 As shown, if the trajectory formed by the human gesture movement is regarded as an antenna array, the spatial direction angle can be calculated as:

[0113]

[0114] Where: Angle[θ,n] is the signal function along the spatial direction θ at the measurement time n; λ is the wavelength; Δd is the spatial distance between consecutive antennas in the array; and S[n+i] is the antenna array.

[0115] At any point in time n, the maximum value of θ in Angle[θ,n] corresponds to the direction of the object's motion. To calculate Angle[θ,n] from the above formula, it is necessary to estimate the spatial distance Δd between consecutive antennas in the simulated array. Δd can be expressed as Δd = vT, where T is the sampling period and v is the speed of motion. Since human motion simulates the antennas in the array, although the exact speed of arm movement is unknown, the range of human arm movement speed is quite narrow. Furthermore, errors in the value of v translate into underestimation or overestimation of the exact direction of the human body and do not hinder subsequent occupancy detection. Therefore, the default value of v = 1 m / s can be selected.

[0116] The smoothed MUSIC (Multiple Signal Classification) algorithm is a technique for estimating the direction of arrival (DOA) of a signal by decomposing the covariance matrix. The MUSIC algorithm achieves high-resolution DOA estimation by analyzing the orthogonality between the signal and noise subspaces. The basic idea is that the eigenvectors of the covariance matrix can be divided into a signal subspace and a noise subspace. The eigenvectors of the signal subspace are related to the direction of arrival of the signal, while the eigenvectors of the noise subspace are orthogonal to these directions. The algorithm steps are as follows:

[0117] ① Construct the covariance matrix R of the received signal, R=E[xx H ].

[0118] ② Calculate the eigenvalues ​​and eigenvectors of the covariance matrix and divide it into signal subspace and noise subspace.

[0119] ③By processing the eigenvectors of the noise subspace, the MUSIC space spectrum function is formed:

[0120]

[0121] Where: W i Is a transformation matrix used to filter the signal; e i is the eigenvector of the noise subspace; a(θ) is the array manifold vector.

[0122] ④ By finding the peak of the MUSIC spectrum, the arrival direction of the signal is estimated, and finally a clear trajectory formed by human movement is obtained in the time direction spectrum. The number of people in the environment is obtained based on the number of trajectories.

[0123] This long-distance wireless sensing method for the unstructured environment of coal mines utilizes periodically operating transmitting and receiving antennas to perform detection that avoids signal interference and ensures the accuracy and reliability of signal reception. After a disaster occurs, it can determine the specific location of trapped personnel based on the amplitude change direction of their gesture signals and the location of the signal source. It also combines the smoothed MUSIC algorithm and the ISAR algorithm to accurately separate and analyze the reflected signals, and then determine the number and relative positions of people in the environment, thereby achieving long-distance perception and accurate positioning of trapped personnel. It can ensure effective signal transmission and detection in complex underground coal mine tunnel environments, and is particularly suitable for rescue detection and locating trapped personnel in underground coal mine environments after disasters.

Claims

1. A long-distance wireless sensing method for unstructured coal mine environments, characterized in that: The non-contact advanced detection system used includes an intelligent detection robot, a receiving antenna, a transmitting antenna and a reflector. A pair of receiving antennas are installed side by side on the intelligent detection robot. The transmitting antenna is a directional transmitting antenna, and there are multiple transmitting antennas. Each transmitting antenna is equipped with a LoRa node. One transmitting antenna is installed on the intelligent detection robot, and the remaining transmitting antennas are installed in the underground coal mine tunnel. The reflector with a metal sheet structure is installed on the wall of the underground coal mine tunnel. The autonomously movable intelligent detection robot includes a signal processing module, and the signal processing module is electrically connected to a pair of receiving antennas and a transmitting antenna on the intelligent detection robot respectively. The long-distance wireless sensing method for unstructured coal mine environments specifically includes the following steps: Step 1, equipment layout: Install transmitting antennas at set intervals along the length of the coal mine tunnel, with the transmitting antennas facing the same direction. Set the transmission cycle and transmission time of each transmitting antenna. After the transmitting antennas are installed, record the installation position and specific working time of each transmitting antenna; After determining the optimal reflective installation position of the reflector, install the reflector on the wall of the coal mine tunnel; Conduct on-site exploration in the coal mine tunnel to determine the location with good signal detection effect, mark and record the specific coordinates and environmental description of each location with good signal detection effect; Step 2, post-disaster detection: After a disaster occurs, the coal mine tunnel layout data with locations with good detection results marked is first input into the intelligent detection robot. The intelligent detection robot is then controlled to enter the coal mine tunnel and go to the marked locations with good detection results to perform signal detection. The intelligent detection robot waits for at least one complete signal transmission cycle at each location with good detection results to receive signals from all transmitting antennas. Step 2-1, determine the location of the trapped person: The signal processing module records and analyzes the received signal to determine whether there is a gesture signal. When a gesture signal is detected, the signal processing module first determines the transmitter of the gesture signal. If the transmitter is a transmitting antenna in the coal mine tunnel, the signal processing module determines that the trapped person is near the transmitting antenna. If the transmitter is the transmitting antenna on the intelligent detection robot, the signal processing module first determines the amplitude direction of the gesture signal. Then, based on the amplitude fluctuation direction of the gesture signal, it determines which side of the intersection the trapped person is located relative to the intelligent detection robot. The details are as follows: Receiving antennas RX1 and RX2 are installed side by side on the intelligent detection robot. When determining the direction of the trapped person based on the amplitude fluctuation direction of the gesture signal, the amplitude change caused by path attenuation is smaller than the change caused by the different reflection areas. When the signal comes from the right corner of the receiving antenna, the reflection area of ​​the signal entering the receiving antenna RX1 from the wall is larger than the reflection area of ​​the signal entering the receiving antenna RX2 from the wall. The signal strength increment of the receiving antenna RX1 is larger than the signal strength increment of the receiving antenna RX2. When processing the LoRa signal, the receiving antennas RX1 and RX2 are used to eliminate the carrier frequency offset and sampling frequency offset. The obtained amplitude is expressed as: Where: amp is the collected amplitude; abs represents the absolute value function; and They are the signals received by the two receiving antennas on the intelligent detection robot; Then, when the signal is reflected from the right corner, there is an 85% probability that The increase is greater than The increase in the amount of , which causes the amp to become larger; when the signal is reflected from the left corner, there is an 85% probability that The increase is less than The increase in , which causes amp to decrease; Therefore, if the amplitude fluctuation direction of the hand signal is upward, it means that the trapped person is at the right corner of the intersection; if the amplitude fluctuation direction of the hand signal is downward, it means that the trapped person is at the left corner of the intersection. If there are both upward and downward amplitude fluctuation directions, it means that there are trapped people at both the left and right corners of the intersection. Step 2-2, determine the number of trapped people: After confirming the existence of trapped people, the reflected signal is processed by eliminating the static signal and direct path signal in the received signal, separating the reflected signals in different directions, and determining the arrival angle of the reflected signal. The position of the trapped people relative to the receiving antenna and the number of trapped people are calculated.

2. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 2-1, the signal processing module uses the singular spectrum analysis method, the multi-scale sliding window fluctuation detection method, and the average gradient judgment method to judge the amplitude trend of the gesture signal; The specific steps of the singular spectrum analysis method are as follows: ① Embed a one-dimensional time series into a high-dimensional space. Given a time series of length N (X1, X2, …, X N ), select a window length L (usually L < N / 2), and construct an L×K trajectory matrix X, where K = N - L + 1. Each column of the trajectory matrix represents a subsequence of the time series, as shown below: X=[X1,X2,…,X K ],X I (x i , x i+1 ,…,x i+L-1 ) T Perform SVD decomposition on the trajectory matrix X and get: Where: f is a nonlinear function; λ i It's XX T The characteristic value of U I and V i are the left and right singular vectors respectively; s is the number of non-zero singular values; αX0 is an additional matrix that increases the degree of freedom; ② According to the size and physical meaning of the eigenvalue, the decomposed components are grouped. Large eigenvalues ​​correspond to trend and periodic components, and small eigenvalues ​​correspond to noise components. ③ Perform inverse transformation on the selected components to reconstruct the time series. By selecting different components, the trend, cycle and noise parts of the time series can be extracted respectively; The specific steps of the multi-scale sliding window fluctuation detection method are as follows: ①Select sliding windows of different sizes for data processing; ② Calculate the degree of fluctuation of the data in each window. If the fluctuation in the window is less than the predetermined threshold, the data in the window is processed as 0; ③ Perform fluctuation detection on each sliding window size separately, as follows: a. Use the maximum window size to process the entire data and obtain the first-level processing results; b. Process the first-level processing results using a medium window size to obtain the second-level processing results; c. Process the second-level processing results using the minimum window size to obtain the final result; The specific steps of the average gradient judgment method are as follows: ① Calculate the gradient of the initial part of each gesture signal amplitude data. The gradient formula is: Where: ΔA is the amplitude difference between adjacent sampling points; Δt is the sampling interval; ②Average the gradient values ​​of each segment of data to obtain the average gradient of the segment of data: Where: Ns is the number of sampling points of this segment of data; ③ Judge the trend of the signal based on the sign and size of the average gradient. If the average gradient is positive, it means that the signal amplitude is increasing; if the average gradient is negative, it means that the signal amplitude is decreasing.

3. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 2-2, when processing the reflected signal by eliminating the static signal and direct path signal in the received signal, the inverse synthetic aperture radar algorithm and smoothed MUSIC algorithm are applied to perform in-depth analysis of the signal; When applying the inverse synthetic aperture radar algorithm to perform in-depth signal analysis, the trajectory formed by the human gesture movement is regarded as the antenna array. The spatial direction angle calculation formula is as follows: Where: Angle[θ,n] is the signal function along the spatial direction θ at the measurement time n; λ is the wavelength; Δd is the spatial distance between consecutive antennas in the array; S[n+i] is the antenna array; Estimate the spatial distance Δd between consecutive antennas in the simulation array. Δd is expressed as Δd = vT, where T is the sampling period and v is the velocity. The default value of v is v = 1 m / s. When applying the smoothed MUSIC algorithm to perform in-depth analysis of a signal, the specific steps are as follows: ① Construct the covariance matrix R of the received signal, R=E[xx H ]; ② Calculate the eigenvalues ​​and eigenvectors of the covariance matrix and divide it into signal subspace and noise subspace; ③By processing the eigenvectors of the noise subspace, the MUSIC space spectrum function is formed: Where: W i Is a transformation matrix used to filter the signal; e i is the eigenvector of the noise subspace; a(θ) is the array manifold vector; ④ By finding the peak of the MUSIC spectrum, the arrival direction of the signal is estimated, and finally a clear trajectory formed by human movement is obtained in the time direction spectrum. The number of people in the environment is obtained based on the number of trajectories.

4. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 2, when the intelligent detection robot goes to the marked locations with good detection effects to perform signal detection, it selects at least two locations with good detection effects at each intersection for signal detection.

5. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 2-1, if the emission source is the transmitting antenna on the intelligent detection robot, repeat the detection three times at the location with good detection effect, and select the case where the detection result is greater than or equal to twice.

6. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 1, when installing transmitting antennas in underground coal mine tunnels, for crossroads and T-junctions, according to the detection range of the receiving antenna on the intelligent detection robot, the geometric center of the intersection is used as the range center, and no transmitting antenna is installed within the set range.

7. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 1, when determining the optimal reflective installation position of the reflector, detection equipment is used to conduct on-site exploration of the underground tunnels of the coal mine, and the optimal reflective installation position of the reflector is determined through experiments and data analysis.

8. The long-distance wireless sensing method for coal mine non-structured environment according to claim 1 is characterized in that: In Step 1, when determining locations with good signal detection effects, for crossroads and T-junctions, multiple locations with good detection effects are determined for each intersection.

Images