A coal mine underground personnel accurate positioning method and system

Abstract

The present application relates to the technical field of positioning measurement, and discloses a kind of coal mine underground personnel accurate positioning method and system, including obtaining the ranging information set between the personnel label to be positioned and positioning base station;Construct the mixed probability model of line-of-sight and non-line-of-sight propagation, the particle filtering algorithm is used to iteratively estimate personnel location, based on the confidence of the positioning result determined in last period, adjust the process noise covariance of particle propagation process;The joint likelihood function of particle under all positioning base stations is calculated using the mixed probability model and the weight is updated;The density clustering algorithm is executed to the particle set whose weight is higher than the preset threshold, and the weighted centroid of the maximum density cluster is output as the personnel positioning coordinates of the current period, and the positioning result confidence is calculated.The present application can accurately represent the ranging error under line-of-sight and non-line-of-sight propagation, and adaptively adjust the process noise using confidence, overcome the particle distribution multi-peak problem caused by NLOS interference, and improve the accuracy of underground personnel positioning.

Description

Technical Field

[0001] This invention relates to the field of positioning and measurement technology, specifically to a method and system for precise positioning of personnel underground in coal mines. Background Technology

[0002] Underground personnel positioning systems in coal mines are crucial for ensuring miners' safety, enabling intelligent mine management, and facilitating emergency rescue command. Ultra-wideband (UWB) technology, due to its high temporal resolution and strong penetration, has been widely applied in this field. However, the underground environment in coal mines is extremely complex, with narrow tunnels, numerous support structures, and severe electromagnetic interference. These factors cause frequent obstruction and reflection of wireless signals during propagation, leading to serious non-line-of-sight (NLOS) propagation problems. NLOS propagation introduces a significant positive bias into UWB ranging, and traditional positioning algorithms (such as least squares or extended Kalman filtering) typically assume that the measurement noise follows a zero-mean Gaussian distribution. Therefore, when NLOS propagation errors exist, the positioning results of these algorithms will exhibit significant deviations.

[0003] There are two main traditional methods to reduce non-line-of-sight (NLOS) propagation errors: The first is to identify and eliminate NLOS measurements. However, in underground environments where NLOS is prevalent, this method may lead to positioning failures due to the removal of too much valid information. The second method attempts to model and compensate for NLOS errors, for example, by constructing a specific probability density function to describe its distribution and combining it with particle filtering (PF) algorithms to handle non-Gaussian and nonlinear problems. However, existing particle filtering schemes still have the following drawbacks: First, their process noise model is fixed, making it difficult to match the switching between stationary and moving states of personnel, easily causing particle set degradation or tracking delays; second, their observation model often oversimplifies the representation of ranging errors, affecting the accuracy of particle weights; finally, the method of estimating position by weighted averaging of all particles is highly susceptible to interference from abnormal particle swarms when NLOS interference causes the particle distribution to exhibit multimodality. These drawbacks result in low positioning accuracy for personnel in coal mines. Summary of the Invention

[0004] This invention provides a method and system for precise positioning of personnel underground in coal mines to solve the problem of low positioning accuracy of personnel underground in coal mines in the prior art.

[0005] In a first aspect, the method for precise positioning of personnel underground in coal mines according to the present invention includes the following steps:

[0006] Obtain a set of ranging information between the tag of the person to be located and at least three positioning base stations. The set of ranging information includes the original distance measurement values ​​and channel impulse response characteristic parameters.

[0007] Based on the channel impulse response characteristic parameters and the preset underground environment map information, a hybrid probability model of line-of-sight and non-line-of-sight propagation is constructed. The ranging error of line-of-sight propagation follows a zero-mean Gaussian distribution, and the variance of the ranging error of line-of-sight propagation is determined by the ratio of the first path amplitude to the average noise power. The ranging error of non-line-of-sight propagation follows a skewed normal distribution, and the position, scale, and shape parameters of non-line-of-sight propagation are jointly determined by the geometric structure information of the underground roadway and the movement state of personnel.

[0008] The particle filtering algorithm is used to iteratively estimate the location of personnel. In each iteration cycle, the process noise covariance of the particle propagation process is adjusted based on the confidence of the positioning result determined in the previous cycle. For the movement state of the personnel represented by each particle, the joint likelihood function of the particle under all positioning base stations is calculated using a mixture probability model, and the weights are updated according to the joint likelihood function.

[0009] For particle sets with weights higher than a preset threshold, a density clustering algorithm is executed. The weighted centroid of the highest density cluster is used as the personnel positioning coordinates for the current period. The confidence level of the positioning result is calculated based on the compactness of the particle distribution within the highest density cluster.

[0010] Preferably, the ranging error of the line-of-sight propagation follows a zero-mean Gaussian distribution, and the variance of the ranging error is determined by the ratio of the first radial amplitude to the average noise power, including:

[0011] Extracting the first path amplitude from the channel impulse response and average noise power The variance of the line-of-sight propagation ranging error is calculated using the following formula. :

[0012] ;

[0013] Where C is an empirical constant determined based on experimental data.

[0014] Preferably, the ranging error of the non-line-of-sight propagation follows a skewed normal distribution, and the non-line-of-sight propagation position, scale, and shape parameters are jointly determined by the underground roadway geometry information and the personnel movement status, including:

[0015] By performing a ray tracing algorithm on the downhole environment map information, the extra path length of signal propagation is calculated, and this extra path length is used as a location parameter of a skewed normal distribution. Based on the type of roadway where the personnel are located, the scale parameters are retrieved from the preset roadway-scale parameter mapping table. Based on the estimated personnel movement speed from the previous period, the shape parameters are calculated using the following formula. :

[0016] ;

[0017] in, and For preset coefficients, The estimated speed of personnel movement for the previous period.

[0018] Preferably, adjusting the process noise covariance of the particle propagation process based on the confidence level of the positioning result determined in the previous cycle includes:

[0019] The process noise covariance of particle propagation is adjusted using the following formula:

[0020] ;

[0021] in, Let $\mathbf{k}$ be the process noise covariance for the $k$-th iteration period. The preset maximum process noise covariance, The preset minimum process noise covariance, The preset attenuation coefficient, The confidence level of the localization result in the (k-1)th iteration cycle.

[0022] Preferably, the step of calculating the joint likelihood function of the particle across all positioning base stations using a hybrid probability model includes:

[0023] For any particle, calculate the distance between the particle and each positioning base station; compare the distance with the original distance measurement value of the corresponding positioning base station to obtain the ranging error; substitute the ranging error into the mixed probability model to obtain the likelihood function value of the particle under the corresponding positioning base station; multiply the likelihood function values ​​of the particle under all N positioning base stations to obtain the joint likelihood function of the particle. ;

[0024] ;

[0025] in, Let x represent the ranging information of the i-th positioning base station, and x represent the particle state. Let N be the likelihood function value of the particle at the i-th positioning base station, and N be the total number of positioning base stations.

[0026] Preferably, the step of performing a density clustering algorithm on particle sets with weights higher than a preset threshold includes:

[0027] The preset threshold for particle weight is set to 1.5 / M, where M is the total number of particles; all particles with weights higher than the preset threshold are selected; the DBSCAN algorithm is used to cluster the selected particle set.

[0028] Preferably, the clustering neighborhood radius Eps is set to 0.5 meters, and the minimum number of particles in the neighborhood MinPts is set to 5.

[0029] Preferably, the compactness is calculated in the following way:

[0030] Calculate the Euclidean distance from each particle within the maximum density cluster to the weighted centroid; then take a weighted average of all Euclidean distances to obtain the average discrete distance. Average Discrete Distance The smaller the value, the more compact the particle distribution within the highest density cluster.

[0031] Preferably, the confidence level of the positioning result is calculated using the following formula:

[0032] ;

[0033] in, To determine the confidence level of the location results, The preset adjustment coefficient, is the average discrete distance.

[0034] Secondly, the precise positioning system for underground personnel in coal mines of the present invention includes a memory and a processor. The memory stores computer instructions, and when the processor executes the computer instructions, it implements the aforementioned precise positioning method for underground personnel in coal mines.

[0035] The beneficial effects of this invention are as follows: 1) By constructing a hybrid probability model that reflects the characteristics of channel impulse response, the geometry of underground roadways, and the movement state of personnel, the ranging error under line-of-sight and non-line-of-sight propagation is represented. 2) The confidence level of the positioning result is fed back to adjust process noise, enabling it to adapt to the switching between different states such as stationary and moving personnel. 3) By using density clustering of high-weight particles to determine the location, the positioning deviation problem caused by the multi-peaked particle distribution under severe NLOS interference is overcome, and the positioning result can be extracted from the contaminated particle set. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the method for precise positioning of personnel underground in coal mines, as provided in an embodiment of the present invention. Detailed Implementation

[0037] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0038] like Figure 1 As shown, an embodiment of the precise positioning method for underground personnel in coal mines provided by the present invention includes the following steps:

[0039] S1, obtain a set of ranging information between the tag of the person to be located and at least three positioning base stations. The set of ranging information includes the original distance measurement value and channel impulse response characteristic parameters.

[0040] Specifically, UWB tags worn by miners in the mine conduct bidirectional ranging communication with at least three UWB positioning base stations deployed on the walls of the underground tunnels. The raw distance measurement value is calculated by multiplying the signal transmission time by the speed of light. During each ranging measurement, the receiver of the UWB chip records detailed channel impulse response data. Subsequently, a series of key characteristic parameters are extracted from this data, such as the first path amplitude, total received signal power, Rice K-factor, and root mean square delay spread. These parameters together form the ranging information set, providing a basis for subsequent channel state judgment and ranging error modeling.

[0041] S2. Based on the channel impulse response characteristic parameters and the preset underground environment map information, a hybrid probability model of line-of-sight and non-line-of-sight propagation is constructed. The ranging error of line-of-sight propagation follows a zero-mean Gaussian distribution, and the variance of the ranging error of line-of-sight propagation is determined by the ratio of the first path amplitude to the average noise power. The ranging error of non-line-of-sight propagation follows a skewed normal distribution, and the position, scale, and shape parameters of non-line-of-sight propagation are jointly determined by the geometric structure information of the underground roadway and the movement state of personnel.

[0042] Specifically, firstly, using an offline-collected dataset, a classifier based on channel impulse response feature parameters, such as a support vector machine, is trained to distinguish between line-of-sight (LAS) and non-line-of-sight (NOS) propagation. After training, this classifier can estimate the LAS probability of the current channel in real time during each ranging measurement. Based on this, a probability density function for the ranging error is constructed. In a specific embodiment, this probability density function is expressed as a weighted sum of the Gaussian distribution corresponding to LAS propagation and the skewed normal distribution corresponding to NOS propagation.

[0043] In an optional embodiment, the ranging error of the line-of-sight propagation follows a zero-mean Gaussian distribution, and the variance of the ranging error is determined by the ratio of the first radial amplitude to the average noise power, including:

[0044] Extracting the first path amplitude from the channel impulse response and average noise power The variance of the line-of-sight propagation ranging error is calculated using the following formula. :

[0045] ;

[0046] Where C is an empirical constant determined based on experimental data.

[0047] For example, an empirical constant C, pre-calibrated through experiments, is first used. Assuming the value of C is 0.02, this constant is substituted into the formula to calculate the variance. If the variance of a measurement is calculated to be 0.00011, this variance is used to construct a Gaussian distribution model with a mean of zero. This model represents the ranging error characteristics under line-of-sight propagation conditions and provides an accurate probability assessment of measurement information from the line-of-sight path in subsequent particle filter weight calculations.

[0048] In an optional embodiment, the ranging error of the non-line-of-sight propagation follows a skewed normal distribution, and the non-line-of-sight propagation position, scale, and shape parameters are jointly determined by the underground roadway geometry information and the personnel movement status, including:

[0049] By performing a ray tracing algorithm on the downhole environment map information, the extra path length of signal propagation is calculated, and this extra path length is used as a location parameter of a skewed normal distribution. Based on the type of roadway where the personnel are located, the scale parameters are retrieved from the preset roadway-scale parameter mapping table. Based on the estimated personnel movement speed from the previous period, the shape parameters are calculated using the following formula. :

[0050] ;

[0051] in, and For preset coefficients, The estimated speed of personnel movement for the previous period.

[0052] For example, first, a pre-set two-dimensional map of the underground tunnel is loaded. Then, a ray tracing algorithm is used to simulate the propagation path of the signal from the base station location (e.g., 5, 10) to the current location (e.g., 20, 18) represented by the particle. The algorithm detects obstacles blocking the line-of-sight path. Calculations show that the straight-line distance between the two points is 18.0 meters, while the shortest non-straight-line propagation path is 19.5 meters. The difference, 1.5 meters, is determined as the additional path length for signal propagation and used as a position parameter for a skewed normal distribution. Suppose that based on the particle's position (20, 18), the map information is queried, and it is determined that the particle is currently in the "Side Channel" type. Next, the corresponding scale parameters are obtained by looking up this channel type in the preset "Channel-Scale Parameter Mapping Table". The value is 1.2. Simultaneously, the estimated movement speed v of the person, obtained from the positioning results of the previous cycle, is 1.1 m / s. Using the formula... The shape parameters are calculated. The value is 3.25. Ultimately, a skewed normal distribution is defined by the position, scale, and shape parameters, which is used to describe the ranging error in this specific non-line-of-sight scenario.

[0053] S3. The particle filtering algorithm is used to iteratively estimate the position of the personnel. In each iteration cycle, the process noise covariance of the particle propagation process is adjusted based on the confidence of the positioning result determined in the previous cycle. For the movement state of the personnel represented by each particle, the joint likelihood function of the particle under all positioning base stations is calculated using a mixture probability model, and the weights are updated according to the joint likelihood function.

[0054] In an optional embodiment, adjusting the process noise covariance of the particle propagation process based on the confidence level of the positioning result determined in the previous cycle includes:

[0055] The process noise covariance of particle propagation is adjusted using the following formula:

[0056] ;

[0057] in, Let $\mathbf{k}$ be the process noise covariance for the $k$-th iteration period. The preset maximum process noise covariance, The preset minimum process noise covariance, The preset attenuation coefficient, The confidence level of the localization result in the (k-1)th iteration cycle.

[0058] For example, assume the maximum process noise covariance The minimum process noise covariance is 0.6. The attenuation coefficient is 0.1. Set to 4.0. At the start of the k-th iteration, obtain the confidence score of the localization result output from the (k-1)-th iteration. Assuming the localization result of the (k-1)th iteration is reliable, The value is 0.95. Substituting this value into the formula for calculation, the process noise covariance in the k-th iteration is... Approximately 0.113. Since the confidence level of the localization result in the (k-1)th iteration is very high, the calculated... The value is close to the preset minimum value This results in a smaller particle diffusion range in the k-th iteration, indicating greater confidence in the localization result of the (k-1)-th iteration. Conversely, if the confidence level of the (k-1)-th iteration is very low, for example... =0.1, then The calculated result is approximately 0.502, which is close to the maximum value. This will lead to a more extensive exploration of the state space by the particles to accommodate the uncertainty of the localization results.

[0059] In an optional embodiment, calculating the joint likelihood function of the particle across all positioning base stations using a hybrid probability model includes:

[0060] For any particle, calculate the distance between the particle and each positioning base station; compare the distance with the original distance measurement value of the corresponding positioning base station to obtain the ranging error; substitute the ranging error into the mixed probability model to obtain the likelihood function value of the particle under the corresponding positioning base station; multiply the likelihood function values ​​of the particle under all N positioning base stations to obtain the joint likelihood function of the particle. ;

[0061] ;

[0062] in, Let x represent the ranging information of the i-th positioning base station, and x represent the particle state. Let N be the likelihood function value of the particle at the i-th positioning base station, and N be the total number of positioning base stations.

[0063] For example, let's illustrate the calculation process using a single particle: First, initialize a set of particles, each representing a possible person's position and velocity state. Based on a preset motion model (such as a constant velocity model), predict the state of each particle. Assume that the predicted position coordinates of one particle x are (12, 30). There are two positioning base stations BS1 and BS2 in the current environment, located at (10, 35) and (15, 28) respectively. The original distance measurements of the base stations are... =5.5m and =3.2m. The geometric distance from the particle to each base station was calculated. and , ≈5.39m; ≈3.61m. Then, calculate the ranging error. and ,get =0.11m, =-0.41m. Simultaneously, the system estimates a line-of-sight probability P for each base station link. LOS Assume P LOS1 =0.9, P LOS2 =0.2. Distance measurement error and Substituting each into the respective link's hybrid probability model yields:

[0064] ,

[0065] ,

[0066] The likelihood function values ​​of the particle at the two base stations were obtained by calculation. =0.75, =0.22. Multiplying the likelihood function values ​​of the two base stations, we obtain the joint likelihood function L=0.165 for the particle, and use this value as the unnormalized weight of the particle.

[0067] S4 executes a density clustering algorithm on the particle set with weights higher than a preset threshold, outputs the weighted centroid of the maximum density cluster as the personnel positioning coordinates for the current period, and calculates the confidence level of the positioning result based on the compactness of the particle distribution within the maximum density cluster.

[0068] In an optional embodiment, performing a density clustering algorithm on a set of particles with weights higher than a preset threshold includes:

[0069] A preset threshold for particle weights is set to 1.5 / M, where M is the total number of particles. All particles with weights higher than the preset threshold are selected. The DBSCAN algorithm is used to cluster the selected particle set. The cluster neighborhood radius Eps is set to 0.5 meters, and the minimum number of particles in the neighborhood MinPts is set to 5.

[0070] For example, assuming the total number of particles M is 3000, the preset threshold for particle weights is calculated to be 0.0005. The 3000 particles are iterated through, and the normalized weight value of each particle is compared to 0.0005. For instance, after filtering, 450 particles have weights higher than the preset threshold; these 450 particles are selected into a high-weight particle set. Then, the DBSCAN clustering algorithm is executed on this set of 450 high-weight particles. The algorithm parameters are preset as follows: neighborhood radius Eps = 0.5 meters, minimum number of particles in the neighborhood MinPts = 5. The algorithm randomly selects a particle p from the set and checks how many other particles are contained within a 0.5-meter radius of that particle. If the number of particles is ≥ 5, particle p is marked as the core point, and a new cluster is created. All particles in that neighborhood are added to this cluster, and so on, continuously expanding the cluster boundary from newly added particles until no more new particles can be absorbed. This process is repeated until all particles have been visited. After clustering, one or more particle clusters may be formed. For example, the final result is a main cluster containing 410 particles, a secondary cluster containing 25 particles, and the remaining 15 particles are marked as noise points because they are not assigned to any cluster.

[0071] In an optional embodiment, the compactness is calculated in the following manner:

[0072] Calculate the Euclidean distance from each particle within the maximum density cluster to the weighted centroid; then take a weighted average of all Euclidean distances to obtain the average discrete distance. Average Discrete Distance The smaller the value, the more compact the particle distribution within the highest density cluster.

[0073] The confidence level of the location result is calculated using the following formula:

[0074] ;

[0075] in, To determine the confidence level of the location results, The preset adjustment coefficient, is the average discrete distance.

[0076] For example, after the clustering step, the cluster containing the largest number of particles is identified as the maximum density cluster and considered as the effective particle swarm. The weighted centroid coordinates of this cluster are calculated, and these coordinates are output as the personnel positioning coordinates for the current period. For example, by taking a weighted average of the x and y coordinates of the 410 particles within the cluster, the centroid position is obtained as (45.3, 62.1). The average discrete distance is then calculated. Assuming the calculated It is 0.35 meters. Assuming a preset adjustment coefficient... The confidence level of the location result is calculated using the formula, which is set to 3. The confidence level is 0.35. This positioning result confidence score will be used in the next iteration. The smaller the value, the more compact the particle distribution, and the higher the calculated confidence level.

[0077] The implementation principle of the precise positioning method for underground personnel in coal mines according to this invention is as follows: This invention constructs a hybrid probability model that integrates channel impulse response characteristics, underground roadway geometry, and personnel movement status, enabling it to accurately represent ranging errors under line-of-sight and non-line-of-sight propagation. Furthermore, a feedback mechanism for the confidence level of the positioning result is introduced, using this confidence level to adaptively adjust process noise, thereby better accommodating the switching between different states such as stationary and moving personnel. In addition, density clustering of high-weighted particles is used to determine the final location, effectively overcoming the multi-peak particle distribution problem caused by severe NLOS interference, and robustly extracting the positioning result from the contaminated particle set.

[0078] An embodiment of the precise positioning system for underground personnel in coal mines provided by the present invention includes a memory and a processor. The memory stores computer instructions, and when the processor executes the computer instructions, it implements the precise positioning method for underground personnel in coal mines as described in the above embodiment.

[0079] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for precise positioning of personnel underground in coal mines, characterized in that, Includes the following steps: Obtain a set of ranging information between the tag of the person to be located and at least three positioning base stations. The set of ranging information includes the original distance measurement values ​​and channel impulse response characteristic parameters. Based on the channel impulse response characteristic parameters and the preset underground environment map information, a hybrid probability model of line-of-sight and non-line-of-sight propagation is constructed. The ranging error of line-of-sight propagation follows a zero-mean Gaussian distribution, and the variance of the ranging error of line-of-sight propagation is determined by the ratio of the first path amplitude to the average noise power. The ranging error of non-line-of-sight propagation follows a skewed normal distribution, and the position, scale, and shape parameters of non-line-of-sight propagation are jointly determined by the geometric structure information of the underground roadway and the movement state of personnel. The particle filtering algorithm is used to iteratively estimate the location of personnel. In each iteration cycle, the process noise covariance of the particle propagation process is adjusted based on the confidence of the positioning result determined in the previous cycle. For the movement state of the personnel represented by each particle, the joint likelihood function of the particle under all positioning base stations is calculated using a mixture probability model, and the weights are updated according to the joint likelihood function. For particle sets with weights higher than a preset threshold, a density clustering algorithm is executed. The weighted centroid of the highest density cluster is used as the personnel positioning coordinates for the current period. The confidence level of the positioning result is calculated based on the compactness of the particle distribution within the highest density cluster.

2. The method for precise positioning of personnel underground in coal mines according to claim 1, characterized in that, The ranging error of the line-of-sight propagation follows a zero-mean Gaussian distribution, and the variance of the ranging error is determined by the ratio of the first radial amplitude to the average noise power, including: Extracting the first path amplitude from the channel impulse response and average noise power The variance of the line-of-sight propagation ranging error is calculated using the following formula. : ； Where C is an empirical constant determined based on experimental data.

3. The method for precise positioning of personnel underground in coal mines according to claim 1, characterized in that, The ranging error of the non-line-of-sight propagation follows a skewed normal distribution. The position, scale, and shape parameters of the non-line-of-sight propagation are jointly determined by the geometric structure information of the underground roadway and the movement state of the personnel, including: By performing a ray tracing algorithm on the downhole environment map information, the extra path length of signal propagation is calculated, and this extra path length is used as a location parameter of a skewed normal distribution. Based on the type of roadway where the personnel are located, the scale parameters are retrieved from the preset roadway-scale parameter mapping table. Based on the estimated personnel movement speed from the previous period, the shape parameters are calculated using the following formula. : ； in, and For preset coefficients, The estimated speed of personnel movement for the previous period.

4. The method for precise positioning of personnel underground in coal mines according to claim 1, characterized in that, The adjustment of the process noise covariance of the particle propagation process based on the confidence level of the positioning result determined in the previous cycle includes: The process noise covariance of particle propagation is adjusted using the following formula: ； in, Let $\mathbf{k}$ be the process noise covariance for the $k$-th iteration period. The preset maximum process noise covariance, The preset minimum process noise covariance, The preset attenuation coefficient, The confidence level of the localization result in the (k-1)th iteration cycle.

5. The method for precise positioning of personnel underground in coal mines according to claim 1, characterized in that, The calculation of the joint likelihood function of the particle across all positioning base stations using a hybrid probability model includes: For any particle, calculate the distance between the particle and each positioning base station; compare the distance with the original distance measurement value of the corresponding positioning base station to obtain the ranging error; substitute the ranging error into the mixed probability model to obtain the likelihood function value of the particle under the corresponding positioning base station; multiply the likelihood function values ​​of the particle under all N positioning base stations to obtain the joint likelihood function of the particle. ; ； in, Let x represent the ranging information of the i-th positioning base station, and x represent the particle state. Let N be the likelihood function value of the particle at the i-th positioning base station, and N be the total number of positioning base stations.

6. The method for precise positioning of personnel underground in coal mines according to claim 1, characterized in that, The process of performing density clustering algorithm on particle sets with weights higher than a preset threshold includes: The preset threshold for particle weight is set to 1.5 / M, where M is the total number of particles; all particles with weights higher than the preset threshold are selected; the DBSCAN algorithm is used to cluster the selected particle set.

7. The method for precise positioning of personnel underground in coal mines according to claim 6, characterized in that, The cluster neighborhood radius Eps is set to 0.5 meters, and the minimum number of particles in the neighborhood MinPts is set to 5.

8. The method for precise positioning of personnel underground in coal mines according to claim 1, characterized in that, The compactness is calculated in the following way: Calculate the Euclidean distance from each particle within the maximum density cluster to the weighted centroid; then take a weighted average of all Euclidean distances to obtain the average discrete distance. Mean Discrete Distance The smaller the value, the more compact the particle distribution within the highest density cluster.

9. The method for precise positioning of personnel underground in coal mines according to claim 8, characterized in that, The confidence level of the location result is calculated using the following formula: ； in, To determine the confidence level of the location results, The preset adjustment coefficient, is the average discrete distance.

10. A precise positioning system for personnel underground in a coal mine, characterized in that, It includes a memory and a processor. The memory stores computer instructions. When the processor executes the computer instructions, it implements the precise positioning method for underground personnel in coal mines as described in any one of claims 1-9.

Images