An indoor positioning method, apparatus, system, and storage medium

Abstract

This invention provides an indoor positioning method, device, system, and storage medium, belonging to the field of navigation and positioning. The method includes: S1: obtaining raw angle data from pre-deployed Bluetooth base stations, and constructing a target node feature matrix and a target adjacency matrix based on the raw angle data; S2: constructing a training model, and training the training model based on the target node feature matrix and the target adjacency matrix to obtain a predicted coordinate matrix; S3: analyzing the loss function between the predicted coordinate matrix and the actual coordinate data to obtain a positioning model; S4: using the positioning model to locate the angle data to be located to obtain the indoor positioning result. This invention suppresses the influence of NLOS on positioning and also suppresses noise caused by non-line-of-sight, while having the advantage of high positioning accuracy.

Description

Technical Field

[0001] This invention relates primarily to the field of navigation and positioning technology, specifically to an indoor positioning method, device, system, and storage medium. Background Technology

[0002] Indoor positioning is the process of acquiring the location of a device or user in an indoor environment. Indoor positioning accuracy is severely compromised. Furthermore, in modern society, people spend a significant amount of time indoors each day. To meet people's needs for learning, working, and living, improving the performance of indoor positioning is essential. Currently, commonly used indoor positioning technologies include Bluetooth, UWB, and Wi-Fi, among others. Bluetooth positioning is divided into traditional Bluetooth positioning and Bluetooth AOA positioning. Traditional Bluetooth positioning includes RSSI and fingerprint positioning methods, both of which suffer from low positioning accuracy. Wi-Fi positioning also suffers from low accuracy. UWB positioning offers high accuracy but is costly and requires complex deployment. Indoor positioning using Bluetooth AOA offers significant advantages in terms of availability, energy efficiency, signal reception range, latency, and scalability, in addition to cost and accuracy. While Bluetooth AOA achieves very high accuracy under ideal conditions, in actual indoor positioning, obstacles (tables, chairs, pedestrians, etc.) create a non-line-of-sight (NLOS) state between the base station and the target node, severely impacting positioning accuracy. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide an indoor positioning method, device, system and storage medium to address the shortcomings of the prior art.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: An indoor positioning method, comprising the following steps:

[0005] S1: Obtain raw angle data from pre-deployed Bluetooth base stations, and construct a target node feature matrix and a target adjacency matrix based on the raw angle data;

[0006] S2: Construct a training model, train the training model based on the target node feature matrix and the target adjacency matrix to obtain the predicted coordinate matrix;

[0007] S3: Import the actual coordinate data and analyze the loss function between the predicted coordinate matrix and the actual coordinate data to obtain the localization model;

[0008] S4: Import the angle data to be located, and use the positioning model to locate the angle data to obtain the indoor positioning result.

[0009] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: An indoor positioning device, comprising:

[0010] A matrix construction module is used to obtain raw angle data from pre-deployed Bluetooth base stations and construct a target node feature matrix and a target adjacency matrix based on the raw angle data.

[0011] The training module is used to build a training model and train the training model based on the target node feature matrix and the target adjacency matrix to obtain the prediction coordinate matrix.

[0012] The analysis module is used to import actual coordinate data and analyze the loss function of the predicted coordinate matrix and the actual coordinate data to obtain the localization model;

[0013] The indoor positioning result acquisition module is used to import the angle data to be positioned, and to locate the angle data to be positioned using the positioning model to obtain the indoor positioning result.

[0014] Based on the above-mentioned indoor positioning method, the present invention also provides an indoor positioning system.

[0015] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: an indoor positioning system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the indoor positioning method described above is implemented.

[0016] Based on the above-described indoor positioning method, the present invention also provides a computer-readable storage medium.

[0017] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the indoor positioning method as described above.

[0018] The beneficial effects of this invention are: constructing a target node feature matrix and a target adjacency matrix through the original angle data; training the training model based on the target node feature matrix and the target adjacency matrix to obtain a predicted coordinate matrix; analyzing the loss function between the predicted coordinate matrix and the actual coordinate data to obtain a positioning model; and obtaining indoor positioning results through positioning of the angle data to be positioned using the positioning model. This suppresses the influence of NLOS on positioning and also suppresses noise caused by non-line-of-sight, while having the advantage of high positioning accuracy. Attached Figure Description

[0019] Figure 1 A flowchart illustrating an indoor positioning method provided in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the antenna array in the Bluetooth AOA system provided in an embodiment of the present invention;

[0021] Figure 3 A schematic diagram illustrating the principle of angle of arrival measurement provided in an embodiment of the present invention;

[0022] Figure 4 This is a module block diagram of an indoor positioning device provided in an embodiment of the present invention. Detailed Implementation

[0023] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0024] Figure 1 This is a flowchart illustrating an indoor positioning method provided in an embodiment of the present invention.

[0025] like Figure 1 As shown, an indoor positioning method includes the following steps:

[0026] S1: Obtain raw angle data from pre-deployed Bluetooth base stations, and construct a target node feature matrix and a target adjacency matrix based on the raw angle data;

[0027] S2: Construct a training model, train the training model based on the target node feature matrix and the target adjacency matrix to obtain the predicted coordinate matrix;

[0028] S3: Import the actual coordinate data and analyze the loss function between the predicted coordinate matrix and the actual coordinate data to obtain the localization model;

[0029] S4: Import the angle data to be located, and use the positioning model to locate the angle data to obtain the indoor positioning result.

[0030] It should be understood that a Bluetooth base station (i.e., a pre-deployed Bluetooth base station) is set up in an indoor environment to collect and record a large amount of angle data (i.e., the raw angle data) and the actual coordinate data as training samples.

[0031] It should be understood that the collected coordinate data (i.e., the actual coordinate data) includes the node coordinates (x1, y1), (x2, y2), ..., (x...) of N nodes relative to the base station. N y N ), where the actual coordinates of the nodes are recorded manually.

[0032] In the above embodiments, a target node feature matrix and a target adjacency matrix are constructed using the original angle data. The predicted coordinate matrix is ​​obtained by training the training model based on the target node feature matrix and the target adjacency matrix. The loss function between the predicted coordinate matrix and the actual coordinate data is analyzed to obtain the positioning model. The indoor positioning result is obtained by positioning the angle data to be positioned using the positioning model. This suppresses the influence of NLOS on positioning and also suppresses noise caused by non-line-of-sight, while having the advantage of high positioning accuracy.

[0033] Optionally, as an embodiment of the present invention, the original angle data includes multiple nodes, the original pitch angle corresponding to each node, and the original azimuth angle, and the process of S1 includes:

[0034] Raw angle data is obtained from pre-deployed Bluetooth base stations, and a raw node feature matrix is ​​constructed based on all the raw pitch angles and all the raw azimuth angles.

[0035] The original node feature matrix is ​​normalized to obtain the target node feature matrix, which includes the node feature elevation angle and node feature azimuth angle corresponding to each node.

[0036] Calculate the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the nodes and the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes to obtain the target angle corresponding to each of the nodes, and construct the original adjacency matrix based on all the target angles;

[0037] The original adjacency matrix is ​​transformed to obtain multiple adjacency angles, and a target adjacency matrix is ​​constructed based on all the adjacency angles.

[0038] It should be understood that the collected angle data (i.e., the raw angle data) includes the pitch angles (i.e., the raw pitch angles) θ1, θ2, ..., θ3 of N nodes relative to the base station. N And azimuth angle (i.e., the original azimuth angle) The angle data (i.e., the original angle data) is the angle measured by the base station.

[0039] It should be understood that normalization, or normalization methods, is mainly proposed for the convenience of data processing. By mapping data to the range of 0 to 1, it is processed more conveniently and quickly, and should be categorized under digital signal processing. There are two forms of normalization methods: one is to transform a number into a decimal between (0, 1), and the other is to transform a dimensional expression into a dimensionless expression.

[0040] Specifically, the measured angle data (i.e., the original angle data) is used as node features, and the angle differences between nodes are used to construct an adjacency matrix (i.e., the original adjacency matrix), thus forming graph data and normalizing the corresponding values.

[0041] It should be understood that the N sets of elevation angles θ (i.e., the original elevation angles) and azimuth angles φ (i.e., the original azimuth angles) of the node relative to the Bluetooth base station are used as the node feature matrix X. N×2 (i.e., the original node feature matrix), and normalize it.

[0042] In the above embodiments, a target node feature matrix and a target adjacency matrix are constructed based on the original angle data, providing accurate basic data for subsequent data processing, suppressing the impact of NLOS on positioning, suppressing noise caused by non-line-of-sight, and having the advantage of high positioning accuracy.

[0043] Optionally, as an embodiment of the present invention, the process of calculating the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the nodes and the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes to obtain the target angle corresponding to each of the nodes includes:

[0044] The target angle corresponding to each node is obtained by summing the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the remaining nodes with the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes using the first formula. The first formula is:

[0045]

[0046] Where, Δθ ij Let θ be the target angle between the i-th node and the j-th node. i Let θ be the pitch angle of the node characteristic corresponding to the i-th node. j Let the pitch angle be the feature angle of the j-th node. Let be the characteristic azimuth angle of the i-th node. Let be the characteristic azimuth angle of the j-th node.

[0047] It should be understood that the remaining nodes refer to all nodes other than the current node.

[0048] It should be understood that, after normalization, the sum of the angle differences between the elevation and azimuth angles of each node relative to the other N-1 nodes is used. As the adjacency matrix A N×N (i.e., the original adjacency matrix), where θ i This represents the pitch angle corresponding to the i-th node. This represents the azimuth angle corresponding to the i-th node.

[0049] In the above embodiments, the target angle corresponding to each node is obtained by summing the node characteristic pitch angle and node characteristic azimuth angle corresponding to each node with the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes through the first formula. This provides accurate basic data for subsequent data processing, suppresses the influence of NLOS on positioning, suppresses noise caused by non-line-of-sight, and has the advantage of high positioning accuracy.

[0050] Optionally, as an embodiment of the present invention, the process of transforming the original adjacency matrix to obtain multiple adjacency angles includes:

[0051] The original adjacency matrix is ​​transformed using the second equation to obtain multiple adjacency angles. The second equation is:

[0052]

[0053] Among them, A' ab Let A be the adjacency angle corresponding to the a-th row and b-th column in the original adjacency matrix. ab TH represents the angle difference in the a-th row and b-th column of the original adjacency matrix, and TH is the preset transformation threshold.

[0054] It should be understood that the normalized threshold TH (i.e., the preset conversion threshold) is between 0.04 and 0.08.

[0055] Specifically, a threshold TH is set to determine whether there is an edge between nodes, for the adjacency matrix A. N×N Perform the following processing; use the node's own true coordinates as node features for GCN training, as shown in the following formula:

[0056]

[0057] In the formula, A ij This represents the element in the i-th row and j-th column of matrix A.

[0058] In the above embodiments, multiple adjacency angles are obtained by transforming the original adjacency matrix through the second equation. The additional information of the edges can be used to predict the nodes, which greatly improves the positioning accuracy in non-line-of-sight situations.

[0059] Optionally, as an embodiment of the present invention, the training model includes a graph convolutional neural network, and the process of S2 includes:

[0060] The graph convolutional neural network is constructed, and trained based on the target node feature matrix and the target adjacency matrix to obtain the predicted coordinate matrix, specifically as follows:

[0061] The graph convolutional neural network is trained using the third equation, the target node feature matrix, and the target adjacency matrix to obtain the predicted coordinate matrix. The third equation is:

[0062]

[0063] Where Y is the predicted coordinate matrix, Let X be the degree matrix, σ be the activation function, X be the feature matrix of the target node, and W be the degree matrix. (0) and W (1) All are trainable weights, A is the target adjacency matrix, and I is the identity matrix.

[0064] It should be understood that the neural network parameters are initialized, and the angle information of each node (i.e., the target node feature matrix) is used as input, and the coordinates of the node (i.e., the target adjacency matrix) are used as labels for GCN training.

[0065] Specifically, the two-layer GCN propagation model is as follows:

[0066]

[0067] In the formula, Let A+I be the adjacency matrix (i.e., the target adjacency matrix) and I be the identity matrix. yes The degree matrix, The purpose is to symmetrically normalize the adjacency matrix A, and add the identity matrix to add self-loops; X is the node feature matrix (i.e., the target node feature matrix), which uses pitch and azimuth angles to represent the features of the nodes; σ is the activation function; W represents the trainable weights.

[0068] It should be understood that in a two-layer GCN, the input feature X is an N×2 matrix, the hidden layers are set to 2000, and W is used to obtain an N×2 coordinate output. (0) It should be a 2×N matrix, W (1) It is an N×2 matrix; where the activation function σ is chosen to be the ReLU function with one-sided inhibition.

[0069] Specifically, GCN takes the following form:

[0070]

[0071] In the formula: I is the identity matrix, added to incorporate self-loops, representing the relationships between nodes; H represents the features of each layer. In this invention, the first layer features are the elevation and azimuth angles of the node relative to the base station, and the second layer features are the two-dimensional coordinates of the node; σ is the activation function. This invention uses a one-sided suppression ReLU function, specifically ReLU(x) = max(0,x); W (l) These are trainable parameters, where l represents the number of layers in each layer; yes The degree matrix, specifically in the form of

[0072] In the above embodiments, the graph convolutional neural network is constructed, and the predicted coordinate matrix is ​​obtained by training the graph convolutional neural network according to the target node feature matrix and the target adjacency matrix. This suppresses the influence of NLOS on localization and also suppresses noise caused by non-line-of-sight, while having the advantage of high localization accuracy.

[0073] Optionally, as an embodiment of the present invention, the process of S3 includes:

[0074] Import the actual coordinate data, and use the root mean square error algorithm to calculate the loss function between the predicted coordinate matrix and the actual coordinate data to obtain the loss value;

[0075] Determine whether the loss value is greater than or equal to a preset threshold. If not, update the parameters of the training model based on the loss value to obtain the updated training model and return to S1. If yes, use the training model as the localization model.

[0076] It should be understood that the root mean square error (RMSE) is used as the loss function to train the GCN (i.e., the training model). If the loss function (i.e., the loss value) does not meet expectations, the process returns to S1 to readjust the parameters and train again.

[0077] In the above embodiments, the loss function of the predicted coordinate matrix and the actual coordinate data is analyzed to obtain the positioning model, which suppresses the influence of NLOS on positioning and also suppresses noise caused by non-line-of-sight, while having the advantage of high positioning accuracy.

[0078] Alternatively, as another embodiment of the present invention, such as Figure 2 As shown, the antenna array used in this invention is a 12-antenna rectangular antenna, which is mounted on the receiver. By switching the antennas, the direction to the transmitter can be determined, i.e., the azimuth angle can be determined. Figure 2 In the diagram, each small square represents an antenna, and the antenna array consists of 12 antennas.

[0079] Alternatively, as another embodiment of the present invention, such as Figure 3As shown, the distance between the two antennas in this invention is d, and the other adjacent side of the arrival angle θ is the distance difference of the paths taken by the same signal received by the two antenna elements. It is related to the phase difference ψ between the signals they receive, and the formula can be obtained as follows:

[0080] ψ=[2πdcos(θ)] / λ

[0081] Where λ represents the wavelength of the signal, the expression for the angle of arrival θ is as follows:

[0082]

[0083] Optionally, as another embodiment of the present invention, the present invention uses the Bluetooth AOA method to collect the elevation and azimuth construction map data of the node, establishes a GCN model, trains it, and uses the trained model to predict the position of the point to be measured. The present invention can suppress noise caused by non-line-of-sight and has the advantage of high positioning accuracy.

[0084] Alternatively, as another embodiment of the present invention, the specific steps of the present invention are as follows:

[0085] S1: Deploy a Bluetooth base station in an indoor environment to collect and record a large amount of angle data and actual coordinate data as training samples;

[0086] S2: Use the angle data measured in S1 as node features, construct an adjacency matrix from the angle differences between nodes, build graph data, and normalize the corresponding values.

[0087] S3: Initialize the neural network parameters, using the angle information of each node as input and the coordinates of the node as labels, and perform GCN training;

[0088] S4: Obtain the graph convolutional neural network model, and input the node angle data that needs to be predicted into the model;

[0089] S5: Predict and output coordinate positions.

[0090] Alternatively, as another embodiment of the present invention, compared with multilayer perceptrons (MLP), GNNs can utilize additional edge information to predict nodes. GCNs can provide high-precision positioning very stably. Simulation experiments show that GCNs can significantly improve positioning accuracy in non-line-of-sight situations using Bluetooth AOA positioning.

[0091] Optionally, as another embodiment of the present invention, in actual data acquisition, the pitch angle error is within 5° and the azimuth angle error is within 10° under line-of-sight conditions.

[0092] Figure 4This is a module block diagram of an indoor positioning device provided in an embodiment of the present invention.

[0093] Alternatively, as another embodiment of the present invention, such as Figure 4 As shown, an indoor positioning device includes:

[0094] A matrix construction module is used to obtain raw angle data from pre-deployed Bluetooth base stations and construct a target node feature matrix and a target adjacency matrix based on the raw angle data.

[0095] The training module is used to build a training model and train the training model based on the target node feature matrix and the target adjacency matrix to obtain the prediction coordinate matrix.

[0096] The analysis module is used to import actual coordinate data and analyze the loss function of the predicted coordinate matrix and the actual coordinate data to obtain the localization model;

[0097] The indoor positioning result acquisition module is used to import the angle data to be positioned, and to locate the angle data to be positioned using the positioning model to obtain the indoor positioning result.

[0098] Optionally, as an embodiment of the present invention, the original angle data includes multiple nodes, the original pitch angle and the original azimuth angle corresponding to each node, and the process of the matrix construction module includes:

[0099] Raw angle data is obtained from pre-deployed Bluetooth base stations, and a raw node feature matrix is ​​constructed based on all the raw pitch angles and all the raw azimuth angles.

[0100] The original node feature matrix is ​​normalized to obtain the target node feature matrix, which includes the node feature elevation angle and node feature azimuth angle corresponding to each node.

[0101] Calculate the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the nodes and the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes to obtain the target angle corresponding to each of the nodes, and construct the original adjacency matrix based on all the target angles;

[0102] The original adjacency matrix is ​​transformed to obtain multiple adjacency angles, and a target adjacency matrix is ​​constructed based on all the adjacency angles.

[0103] Optionally, another embodiment of the present invention provides an indoor positioning system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the indoor positioning method as described above. This system can be a computer or similar system.

[0104] Optionally, another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the indoor positioning method as described above.

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0107] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0108] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0109] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0110] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0111] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An indoor positioning method, characterized in that, Includes the following steps: S1: Obtain raw angle data from pre-deployed Bluetooth base stations, and construct a target node feature matrix and a target adjacency matrix based on the raw angle data; S2: Construct a training model, train the training model based on the target node feature matrix and the target adjacency matrix to obtain the predicted coordinate matrix; S3: Import the actual coordinate data and analyze the loss function between the predicted coordinate matrix and the actual coordinate data to obtain the localization model; S4: Import the angle data to be located, and use the positioning model to locate the angle data to obtain the indoor positioning result.

2. The indoor positioning method according to claim 1, characterized in that, The raw angle data includes multiple nodes, the raw pitch angle corresponding to each node, and the raw azimuth angle. The process S1 includes: Raw angle data is obtained from pre-deployed Bluetooth base stations, and a raw node feature matrix is ​​constructed based on all the raw pitch angles and all the raw azimuth angles. The original node feature matrix is ​​normalized to obtain the target node feature matrix, which includes the node feature elevation angle and node feature azimuth angle corresponding to each node. Calculate the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the nodes and the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes to obtain the target angle corresponding to each of the nodes, and construct the original adjacency matrix based on all the target angles; The original adjacency matrix is ​​transformed to obtain multiple adjacency angles, and a target adjacency matrix is ​​constructed based on all the adjacency angles.

3. The indoor positioning method according to claim 2, characterized in that, The process of calculating the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the remaining nodes with the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes to obtain the target angle corresponding to each of the nodes includes: The target angle corresponding to each node is obtained by summing the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the remaining nodes with the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes using the first formula. The first formula is: Where, Δθ ij Let θ be the target angle between the i-th node and the j-th node. i Let θ be the pitch angle of the node characteristic corresponding to the i-th node. j Let the pitch angle be the feature angle of the j-th node. Let be the characteristic azimuth angle of the i-th node. Let be the characteristic azimuth angle of the j-th node.

4. The indoor positioning method according to claim 2, characterized in that, The process of transforming the original adjacency matrix to obtain multiple adjacency angles includes: The original adjacency matrix is ​​transformed using the second equation to obtain multiple adjacency angles. The second equation is: Among them, A' ab Let A be the adjacency angle corresponding to the a-th row and b-th column in the original adjacency matrix. ab TH represents the angle difference in the a-th row and b-th column of the original adjacency matrix, and TH is the preset transformation threshold.

5. The indoor positioning method according to claim 2, characterized in that, The training model includes a graph convolutional neural network, and the process of S2 includes: The graph convolutional neural network is constructed, and trained based on the target node feature matrix and the target adjacency matrix to obtain the predicted coordinate matrix, specifically as follows: The graph convolutional neural network is trained using the third equation, the target node feature matrix, and the target adjacency matrix to obtain the predicted coordinate matrix. The third equation is: Where Y is the predicted coordinate matrix, Let X be the degree matrix, σ be the activation function, X be the feature matrix of the target node, and W be the degree matrix. (0) and W (1) All are trainable weights, A is the target adjacency matrix, and I is the identity matrix.

6. The indoor positioning method according to claim 1, characterized in that, The process S3 includes: Import the actual coordinate data, and use the root mean square error algorithm to calculate the loss function between the predicted coordinate matrix and the actual coordinate data to obtain the loss value; Determine whether the loss value is greater than or equal to a preset threshold. If not, update the parameters of the training model based on the loss value to obtain the updated training model and return to S1. If yes, use the training model as the localization model.

7. An indoor positioning device, characterized in that, include: A matrix construction module is used to obtain raw angle data from pre-deployed Bluetooth base stations and construct a target node feature matrix and a target adjacency matrix based on the raw angle data. The training module is used to build a training model and train the training model based on the target node feature matrix and the target adjacency matrix to obtain the prediction coordinate matrix. The analysis module is used to import actual coordinate data and analyze the loss function of the predicted coordinate matrix and the actual coordinate data to obtain the localization model; The indoor positioning result acquisition module is used to import the angle data to be positioned, and to locate the angle data to be positioned using the positioning model to obtain the indoor positioning result.

8. The indoor positioning device according to claim 7, characterized in that, The raw angle data includes multiple nodes, the raw pitch angle corresponding to each node, and the raw azimuth angle. The matrix construction module process includes: Raw angle data is obtained from pre-deployed Bluetooth base stations, and a raw node feature matrix is ​​constructed based on all the raw pitch angles and all the raw azimuth angles. The original node feature matrix is ​​normalized to obtain the target node feature matrix, which includes the node feature elevation angle and node feature azimuth angle corresponding to each node. Calculate the sum of the node characteristic pitch angle and node characteristic azimuth angle corresponding to each of the nodes and the node characteristic pitch angle and node characteristic azimuth angle corresponding to any of the remaining nodes to obtain the target angle corresponding to each of the nodes, and construct the original adjacency matrix based on all the target angles; The original adjacency matrix is ​​transformed to obtain multiple adjacency angles, and a target adjacency matrix is ​​constructed based on all the adjacency angles.

9. An indoor positioning system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the indoor positioning method as described in any one of claims 1 to 6.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the indoor positioning method as described in any one of claims 1 to 6.

Images